JP2008283490A - Moving image encoding device, method and program, and moving image decoding device, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the efficiency and accuracy of encoding and decoding. <P>SOLUTION: This encoding device is provided with: an area dividing means for dividing an input moving image signal composed of time sequences of a frame image signal into a plurality of processing object areas; a motion detecting means for detecting motion vectors of the processing target areas; a prediction signal generating means for generating a prediction signal for the processing object areas; a residual signal generating means for generating a residual signal between the prediction signal and a processing object area signal; and an encoding means for encoding the residual signal and the motion vector and generating compressed data. The prediction signal generating means encodes a motion vector on the basis of similarity of motion vectors of a plurality of blocks present in frames different from each other or on the basis of the similarity of motion vectors, and similarity of motion vectors of a plurality of blocks present in the same frame as the block of an encoding object. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a moving picture coding apparatus, method and program, and a moving picture decoding apparatus, method and program related to motion compensation by a motion vector.

  In order to efficiently transmit and store still images and moving image data, compression coding technology is used. Examples of conventional video coding systems include MPEG1-MPEG4 and ITU (International Telecommunication Union) H.264. 261-H. H.264 is widely used.

  In these encoding methods, encoding / decoding processing is performed after an image to be encoded is divided into a plurality of blocks. MPEG4 and H.264 In H.264, in order to further increase the coding efficiency, adjacent pre-reproduced image signals (reconstructed compressed image data) in the same screen as the target block are used for predictive coding in the screen. After generating the prediction signal, the difference signal obtained by subtracting the prediction signal from the signal of the target block is encoded. For predictive coding between screens, refer to adjacent previously reproduced image signals in a screen different from the target block, perform motion correction, generate a prediction signal, and subtract it from the signal of the target block The signal is encoded (see Non-Patent Document 1).

  In the inter-frame prediction in the above method, when encoding the motion vector of the target block to be used for motion compensation, the intermediate value of the motion vector of the upper, left and upper left blocks in the same frame as the target block is used as the target. A difference value obtained by subtracting the motion vector prediction value from the motion vector of the target block is encoded as the prediction value of the motion vector of the block (see Patent Document 1).

  Also disclosed is a technique for performing differential encoding using a motion vector of a block in a frame different from a frame including the target block as a motion vector predicted value of the target block when calculating a motion vector predicted value of the target block for encoding. (See Non-Patent Document 2).

  In the MPEG-4 encoding method, a method called direct mode that uses a motion vector of a block in a frame different from that of the target block as a motion vector prediction value of the target block is adjacent in the same frame as the target block. A method is used in which a motion vector of a block is used as a motion vector prediction value of a target block. In order to indicate whether or not the direct mode is used in a certain block, an index indicating a prediction mode is used (see Non-Patent Document 3).

Further, the probability of entropy encoding the motion vector difference value based on the difference value between two motion vectors in adjacent frames, using the motion vector of the block in the frame including the target block as the motion vector prediction value of the target block A technique for increasing the coding efficiency by switching the distribution is disclosed (see Non-Patent Document 4).
US Pat. No. 5,905,535 “ITU-T H.264” TELECOMMUNICATION STANDARDIZATION SECTOR OF ITU SERIES H: AUDIOVISUAL AND MULTIMEDIA SYSTEMS Infrastructure of audiovisual services-Coding of moving video `` MOTION COMPENSATION OF MOTION VECTORS '' Yeh, J .; Vetterli, M .; Khansari, M .; Image Processing, 1995. Proceedings., International Conference on, Volume1, 23-26 Oct. 1995 Page (s): 574-577 vol.1, Digital Object Identifier 10.1109 / ICIP.1995.531431 Text of ISO / IEC 14496-2: 2001 (Unifying N2502, N3307, N3056, and N3664) INTERNATIONAL ORGANIZATION FOR STANDARDIZATION ORGANIZATION INTERNATIONAL NORMALIZATION ISO / IEC JTC 1 / SC 29 / WG 11 CODING OF MOVING PICTURES AND AUDIO `` A Spatial and Temporal Motion Vector Coding Algorithm for Low-Bit-Rate Video Coding '' Chen, MC; Willson, AN, Jr .; Image Processing, 1997. Proceedings., International Conference on Volume2, 26-29 Oct. 1997 Page ( s): 791-794 vol.2 Digital Object Identifier 10.1109 / ICIP.1997.638615

  However, in the prior art based on MPEG-4, information for indicating whether or not the motion vector prediction value of the target block is to be determined using the direct mode must be sent, so that the code amount increases accordingly.

  In addition, since the mode information indicating the motion vector prediction method of the target block, the information for specifying the motion vector of the target block, and the motion vector difference value are entropy-coded based on a fixed probability distribution, Therefore, it cannot be said that the coding efficiency and accuracy are good.

  Further, as an example adaptive to changes in the surrounding environment, there is a conventional technique described in Non-Patent Document 3, but attention is paid only to the spatial direction correlation of motion vectors. For this reason, when the motion vector of a block in a frame different from the target block is used as a motion vector prediction value, the efficiency and accuracy of encoding are not good.

  Furthermore, in the prior art, only the motion vector of the block adjacent to the target block in the same frame as the target block, or the motion vector of the block in the same spatial position as the target block in the frame adjacent to the target block is vector-predicted Because it is a value, it is affected by a local motion vector change. For this reason, it is difficult to obtain a predicted value reflecting a large number of changes in the motion vector, and it is easily affected by noise of the motion vector.

  The present invention has been made to solve the above-described problems, and an object thereof is to improve the efficiency and accuracy of encoding and decoding.

  In order to solve the above-described problem, a moving image encoding device according to the present invention includes an area dividing unit that divides an input moving image signal formed of a time sequence of frame image signals into a plurality of processing target areas, and the processing target. Motion detection means for detecting a motion vector of a region, prediction signal generation means for generating a prediction signal for the processing target region, and residual signal generation means for generating a residual signal between the prediction signal and the processing target region signal And encoding means for generating compressed data by encoding the residual signal and the motion vector, wherein the prediction signal generating means is a similarity between motion vectors of a plurality of blocks in different frames, or Encoding the motion vector based on the similarity and the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. And butterflies.

  The moving image encoding apparatus divides an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, detects a motion vector of the processing target region, and outputs a prediction signal for the processing target region And generating a residual signal between the prediction signal and the processing target region signal, and encoding the residual signal and the motion vector to generate compressed data. Then, the prediction signal generating means is similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be encoded. Based on the above, a motion vector is encoded. As a result, encoding that reflects the majority of changes in motion vectors becomes possible, and the efficiency and accuracy of encoding can be improved.

  In the following, making a motion vector of a block in a frame different from the frame including the target block as a motion vector prediction value of the target block is referred to as “predicting from the time direction”, and the block in the frame including the target block Making the motion vector a motion vector prediction value of the target block is called “predicting from the spatial direction”.

  The predictive signal generation means described above is similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be encoded. It is desirable to determine the motion vector prediction value or select a motion vector prediction value determination method based on the above.

  In this case, the similarity of motion vectors of a plurality of blocks in different frames (for example, information indicating similarity such as a difference value, a variance value, a standard deviation), or the similarity and the same block as the target block Based on the similarity of motion vectors of a plurality of blocks, it is possible to determine a prediction value of a motion vector of a target block or to select a prediction value determination method. Thereby, since the prediction method can be determined automatically, it is not necessary to send information indicating the prediction method, and the code amount can be reduced. In addition, when predicting the motion vector prediction value of the target block from the temporal direction, or when predicting from the temporal direction and the spatial direction, a motion vector with high prediction accuracy is selected and used as a prediction value, thereby predicting with high accuracy. be able to.

  In addition, the prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. It is desirable that the predicted value of the motion vector is determined by weighting based on.

  In this case, by using the surrounding motion vectors effectively by the weighting, it is possible to reduce the influence of local motion vector changes and obtain a predicted value reflecting the majority of motion vector changes.

  In addition, the prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The probability distribution used when encoding information for indicating a motion vector prediction value or the probability distribution used when encoding information for indicating a motion vector prediction value determination method is applied. It is desirable to adopt a configuration that determines automatically.

  In this case, when predicting the motion vector prediction value of the target block from the temporal direction or from the temporal direction and the spatial direction, the probability distribution of each index is adapted according to the surrounding environment, and an appropriate code amount is used. Encoding efficiency can be increased by encoding.

  Further, the prediction signal generation means may perform similarity of motion vectors of a plurality of blocks in different frames, or similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. Based on the above, it is desirable to adopt a configuration that adaptively determines the probability distribution used when coding the motion vector difference value.

  In this case, when predicting the motion vector prediction value of the target block from the temporal direction, or predicting from the temporal direction and the spatial direction, the occurrence probability distribution of the motion vector difference value is adapted according to the surrounding environment, and an appropriate code is used. Coding efficiency can be improved by encoding with a quantity.

  The moving image coding apparatus according to the present invention includes an area dividing unit that divides an input moving image signal composed of a time series of frame image signals into a plurality of processing target areas, and motion detection that detects a motion vector of the processing target areas. Means, a prediction signal generating means for generating a prediction signal for the processing target area, a residual signal generating means for generating a residual signal between the prediction signal and the processing target area signal, the residual signal and the motion A plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. The motion vector is encoded based on the similarity of the motion vectors.

  The moving image encoding apparatus divides an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, detects a motion vector of the processing target region, and outputs a prediction signal for the processing target region And generating a residual signal between the prediction signal and the processing target region signal, and encoding the residual signal and the motion vector to generate compressed data. Then, the prediction signal generating means encodes the motion vector based on the similarity of the motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. It is characterized by doing. As a result, encoding that reflects the majority of changes in motion vectors becomes possible, and the efficiency and accuracy of encoding can be improved.

  Note that the “plurality of blocks” may include blocks in a frame that is one or more frames apart from a frame that includes a block to be encoded.

  The prediction signal generation means described above is based on the similarity of motion vectors of a plurality of blocks including blocks separated by one or more blocks from the target block in the same frame as the block to be encoded. It is desirable to adopt a configuration that performs determination or selection of a motion vector prediction value determination method.

  In this case, when predicting the motion vector prediction value of the target block from the spatial direction, it is possible to accurately predict by reflecting a large number of motion vector changes. It is also possible to cancel the influence of noise and local changes.

  In addition, the prediction signal generation unit may calculate a motion vector prediction value by weighting based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adopt a configuration that determines

  In this case, by using the surrounding motion vectors effectively by the weighting, it is possible to reduce the influence of local motion vector changes and obtain a predicted value reflecting the majority of motion vector changes.

  Further, the prediction signal generation means calculates a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adopt a configuration that adaptively determines a probability distribution used when encoding information for indicating or a probability distribution used when encoding information for indicating a motion vector prediction value determination method. .

  In this case, when predicting a motion vector prediction value of a target block from a spatial direction, a probability distribution used when encoding information for indicating a motion vector prediction value or a motion vector prediction value determination method is shown. The coding efficiency can be increased by adapting the probability distribution used when coding the information for coding according to the surrounding environment and coding with an appropriate code amount.

  Further, the prediction signal generation means encodes the motion vector difference value based on the similarity of the motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adopt a configuration that adaptively determines the probability distribution to be used at the time of conversion.

  In this case, when predicting the motion vector prediction value of the target block from the spatial direction, the probability distribution used when encoding the motion vector difference value is adapted according to the surrounding environment and encoded with an appropriate code amount. Thus, encoding efficiency can be increased.

  In order to solve the above-described problem, a moving picture decoding apparatus according to the present invention includes data extraction means for extracting residual signal and motion vector information related to a processing target area from compressed data, and a residual related to the processing target area. A residual signal restoring means for restoring the difference signal to a reproduction residual signal; a prediction signal generating means for generating a prediction signal for the region to be processed; and adding the prediction signal and the reproduction residual signal, Image restoration means for restoring a pixel signal of a region, wherein the prediction signal generation means is similar to motion vectors of a plurality of blocks in different frames or the same frame as the similarity and a block to be decoded The motion vector is decoded based on the similarity of motion vectors of a plurality of blocks in the block.

  According to this moving image decoding apparatus, the residual signal and motion vector information relating to the processing target region are extracted from the compressed data, the residual signal relating to the processing target region is restored to the reproduction residual signal, and the processing By generating a prediction signal for the target region and adding the prediction signal and the reproduction residual signal, the pixel signal of the target region can be restored.

  Then, the prediction signal generating means is similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be decoded. Based on the above, the motion vector is decoded. As a result, decoding that reflects the majority of changes in motion vectors is possible, and the efficiency and accuracy of the decoding process can be improved.

  The predictive signal generation means described above is similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be decoded. It is desirable to determine the motion vector prediction value or select a motion vector prediction value determination method based on the above.

  In this case, since the prediction method can be automatically determined, it is not necessary to input information indicating the prediction method. In addition, when predicting the motion vector prediction value of the target block from the temporal direction, or when predicting from the temporal direction and the spatial direction, a motion vector with high prediction accuracy is selected and used as a prediction value, thereby predicting with high accuracy. be able to.

  In addition, the prediction signal generating unit may be similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be decoded. It is desirable that the predicted value of the motion vector is determined by weighting based on.

  In this case, by using the surrounding motion vectors effectively by the weighting, it is possible to reduce the influence of local motion vector changes and obtain a predicted value reflecting the majority of motion vector changes.

  In addition, the prediction signal generating unit may be similar to motion vectors of a plurality of blocks in different frames, or similarities of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The probability distribution used when decoding the information for indicating the motion vector prediction value or the probability distribution used when decoding the information for indicating the motion vector prediction value determination method is applied based on It is desirable to adopt a configuration that determines automatically.

  In this case, the probability distribution used when decoding the information for indicating the motion vector prediction value or the probability distribution used for decoding the information for indicating the motion vector prediction value determination method is the peripheral environment. By using the same probability distribution as that of the moving picture coding apparatus, data can be extracted.

  Further, the prediction signal generation means may perform similarity of motion vectors of a plurality of blocks in different frames, or similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. Based on the above, it is desirable to adopt a configuration that adaptively determines the probability distribution used when decoding the motion vector difference value.

  In this case, when predicting the motion vector prediction value of the target block from the temporal direction, or predicting from the temporal direction and the spatial direction, the motion vector difference value occurrence probability distribution is adapted according to the surrounding environment, and the video code Data can be extracted by using the same probability distribution as that of the conversion apparatus.

  The moving picture decoding apparatus according to the present invention includes a data extraction unit that extracts residual signal and motion vector information related to a processing target region from compressed data, and restores the residual signal related to the processing target region into a reproduction residual signal. Residual signal restoring means for performing prediction signal generating means for generating a prediction signal for the processing target region, and image restoration for restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal And the prediction signal generating means is based on similarity of motion vectors of a plurality of blocks including a block that is one block or more apart from the target block in the same frame as the block to be decoded. Is decrypted.

  According to this moving image decoding apparatus, the residual signal and motion vector information relating to the processing target region are extracted from the compressed data, the residual signal relating to the processing target region is restored to the reproduction residual signal, and the processing By generating a prediction signal for the target region and adding the prediction signal and the reproduction residual signal, the pixel signal of the target region can be restored.

  Then, the prediction signal generation unit decodes the motion vector based on the similarity of the motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded. It is characterized by that. As a result, decoding that reflects the majority of changes in motion vectors is possible, and the efficiency and accuracy of the decoding process can be improved.

  The prediction signal generation means described above is based on the similarity of motion vectors of a plurality of blocks including blocks separated by one or more blocks from the target block in the same frame as the block to be decoded. It is desirable to adopt a configuration that performs determination or selection of a motion vector prediction value determination method.

  In this case, since the prediction method can be automatically determined, it is not necessary to input information indicating the prediction method. In addition, when predicting the motion vector prediction value of the target block from the temporal direction, or when predicting from the temporal direction and the spatial direction, a motion vector with high prediction accuracy is selected and used as a prediction value, thereby predicting with high accuracy. be able to.

  In addition, the prediction signal generation unit is configured to calculate a motion vector prediction value by weighting based on the similarity of motion vectors of a plurality of blocks including a block that is one block or more apart from the target block in the same frame as the block to be decoded. It is desirable to adopt a configuration that determines

  In this case, by using the surrounding motion vectors effectively by the weighting, it is possible to reduce the influence of local motion vector changes and obtain a predicted value reflecting the majority of motion vector changes.

  Further, the prediction signal generation means calculates the motion vector prediction value based on the similarity of the motion vectors of a plurality of blocks including a block that is one block or more apart from the target block in the same frame as the block to be decoded. It is desirable to adopt a configuration that adaptively determines a probability distribution used when decoding information for indicating or a probability distribution used when decoding information for indicating a motion vector prediction value determination method. .

  In this case, the probability distribution used when decoding the information for indicating the motion vector prediction value or the probability distribution used for decoding the information for indicating the motion vector prediction value determination method is the peripheral environment. By using the same probability distribution as that of the moving picture coding apparatus, data can be extracted.

  Furthermore, the prediction signal generation means decodes the motion vector difference value based on the similarity of the motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to adopt a configuration that adaptively determines the probability distribution to be used at the time of conversion.

  In this case, when predicting the motion vector prediction value of the target block from the spatial direction, by adapting the occurrence probability distribution of the motion vector difference value according to the surrounding environment, and using the same probability distribution as the moving picture encoding device, Data can be extracted.

  The present invention can be described as the invention related to the moving picture encoding apparatus and the invention related to the moving picture decoding apparatus as described above, as well as the invention related to the moving picture encoding method and the moving picture encoding program as described below. The invention according to the present invention, the invention according to the moving picture decoding method, and the invention related to the moving picture decoding program can also be described. The invention related to the moving image encoding method and the invention related to the moving image encoding program are substantially the same invention except for the category different from the invention related to the moving image encoding apparatus, and have the same operations and effects. Play. In addition, the invention related to the moving picture decoding method and the invention related to the moving picture decoding program are substantially the same invention except for the category different from the invention related to the moving picture decoding apparatus, and have the same operations and effects. Play.

  A moving image encoding method according to the present invention is a moving image encoding method executed by a moving image encoding device, and an input moving image signal composed of a time sequence of frame image signals is assigned to a plurality of processing target regions. A region dividing step for dividing, a motion detecting step for detecting a motion vector of the processing target region, a prediction signal generating step for generating a prediction signal for the processing target region, and the remaining of the prediction signal and the processing target region signal A residual signal generating step for generating a difference signal; and an encoding step for encoding the residual signal and the motion vector to generate compressed data. In the prediction signal generating step, the video encoding device includes: The similarity of motion vectors of a plurality of blocks in different frames, or the same frame as the similarity and the block to be encoded. Based on the similarity of motion vectors of a plurality of blocks, characterized by coding the motion vector.

  In the prediction signal generation step in the moving picture coding method, the moving picture coding apparatus is similar to motion vectors of a plurality of blocks in different frames or the similarity and the block to be coded. It is desirable to determine a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks in a frame.

  In the prediction signal generating step in the video encoding method, the video encoding device may use the similarity of motion vectors of a plurality of blocks in different frames, or the similarity and a block to be encoded. It is desirable to determine a motion vector prediction value determination method based on the similarity of motion vectors of a plurality of blocks in the same frame.

  In the prediction signal generating step in the video encoding method, the video encoding device may use the similarity of motion vectors of a plurality of blocks in different frames, or the similarity and a block to be encoded. It is desirable to determine a motion vector prediction value by weighting based on the similarity of motion vectors of a plurality of blocks in the same frame.

  In the prediction signal generating step in the video encoding method, the video encoding device may use the similarity of motion vectors of a plurality of blocks in different frames, or the similarity and a block to be encoded. It is desirable to adaptively determine a probability distribution used when encoding information for indicating a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks in the same frame.

  In the prediction signal generating step in the video encoding method, the video encoding device may use the similarity of motion vectors of a plurality of blocks in different frames, or the similarity and a block to be encoded. It is desirable to adaptively determine the probability distribution used when encoding information for indicating a motion vector prediction value determination method based on the similarity of motion vectors of a plurality of blocks in the same frame.

  In the prediction signal generating step in the video encoding method, the video encoding device may use the similarity of motion vectors of a plurality of blocks in different frames, or the similarity and a block to be encoded. It is desirable to adaptively determine the probability distribution used when coding the motion vector difference value based on the similarity of the motion vectors of a plurality of blocks in the same frame.

  A moving image encoding method according to the present invention is a moving image encoding method executed by a moving image encoding device, and an input moving image signal composed of a time sequence of frame image signals is assigned to a plurality of processing target regions. A region dividing step for dividing, a motion detecting step for detecting a motion vector of the processing target region, a prediction signal generating step for generating a prediction signal for the processing target region, and the remaining of the prediction signal and the processing target region signal A residual signal generating step for generating a difference signal; and an encoding step for encoding the residual signal and the motion vector to generate compressed data. In the prediction signal generating step, the video encoding device includes: Similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded Based on, characterized by coding the motion vector.

  In the prediction signal generating step in the moving image encoding method, the moving image encoding device includes motion vectors of a plurality of blocks including a block separated by at least one block from the target block in the same frame as the block to be encoded. It is desirable to determine the predicted value of the motion vector based on the similarity of.

  Further, in the prediction signal generation step in the video encoding method, the video encoding apparatus includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to determine a motion vector prediction value determination method based on the similarity of motion vectors.

  Further, in the prediction signal generation step in the video encoding method, the video encoding apparatus includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to determine a motion vector prediction value by weighting based on the similarity of motion vectors.

  Further, in the prediction signal generation step in the video encoding method, the video encoding apparatus includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adaptively determine a probability distribution used when encoding information for indicating a motion vector prediction value based on the similarity of motion vectors.

  Further, in the prediction signal generation step in the video encoding method, the video encoding apparatus includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adaptively determine the probability distribution used when encoding information for indicating a motion vector prediction value determination method based on the similarity of motion vectors.

  Further, in the prediction signal generation step in the video encoding method, the video encoding apparatus includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be encoded. It is desirable to adaptively determine the probability distribution used when coding the motion vector difference value based on the similarity of the motion vectors.

  A moving image decoding method according to the present invention is a moving image decoding method executed by a moving image decoding apparatus, and is a data for extracting residual signal and motion vector information relating to a processing target region from compressed data. An extraction step; a residual signal restoration step that restores a residual signal related to the processing target region to a playback residual signal; a prediction signal generation step that generates a prediction signal for the processing target region; the prediction signal and the playback residual An image restoration step of restoring a pixel signal of the target region by adding a difference signal, wherein in the prediction signal generation step, the moving picture decoding device includes motion vectors of a plurality of blocks in different frames; Or the similarity of motion vectors of blocks in the same frame as the block to be decoded. Zui it, characterized by decoding the motion vector.

  In the prediction signal generation step in the moving picture decoding method, the moving picture decoding apparatus is similar to the similarity of motion vectors of a plurality of blocks in different frames or the similarity and the block to be decoded. It is desirable to determine a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks in a frame.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device is configured to detect the similarity of motion vectors of a plurality of blocks in different frames or the similarity and a block to be decoded. It is desirable to determine a motion vector prediction value determination method based on the similarity of motion vectors of a plurality of blocks in the same frame.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device is configured to detect the similarity of motion vectors of a plurality of blocks in different frames or the similarity and a block to be decoded. It is desirable to determine a motion vector prediction value by weighting based on the similarity of motion vectors of a plurality of blocks in the same frame.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device is configured to detect the similarity of motion vectors of a plurality of blocks in different frames or the similarity and a block to be decoded. It is desirable to adaptively determine a probability distribution used when decoding information for indicating a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks in the same frame.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device is configured to detect the similarity of motion vectors of a plurality of blocks in different frames or the similarity and a block to be decoded. It is desirable to adaptively determine a probability distribution used when decoding information for indicating a motion vector prediction value determination method based on the similarity of motion vectors of a plurality of blocks in the same frame.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device is configured to detect the similarity of motion vectors of a plurality of blocks in different frames or the similarity and a block to be decoded. It is desirable to adaptively determine a probability distribution used when decoding a motion vector difference value based on the similarity of motion vectors of a plurality of blocks in the same frame.

  A moving image decoding method according to the present invention is a moving image decoding method executed by a moving image decoding apparatus, and is a data for extracting residual signal and motion vector information relating to a processing target region from compressed data. An extraction step; a residual signal restoration step that restores a residual signal related to the processing target region to a playback residual signal; a prediction signal generation step that generates a prediction signal for the processing target region; the prediction signal and the playback residual An image restoration step of restoring the pixel signal of the target region by adding the difference signal is provided, and in the prediction signal generation step, the video decoding device includes the target block within the same frame as the block to be decoded. The motion vector is decoded based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more And wherein the door.

  In the motion prediction signal generation step in the moving picture decoding method, the moving picture decoding apparatus moves a plurality of blocks including a block separated by at least one block from the target block in the same frame as the block to be decoded. It is desirable to determine a motion vector prediction value based on vector similarity.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to determine a motion vector prediction value determination method based on the similarity of motion vectors.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to determine a motion vector prediction value by weighting based on the similarity of motion vectors.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to adaptively determine a probability distribution used when decoding information for indicating a motion vector prediction value based on the similarity of motion vectors.

  In the motion prediction signal generation step in the video decoding method, the video decoding device includes a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to adaptively determine a probability distribution used when decoding information for indicating a motion vector prediction value determination method based on the similarity of motion vectors.

  Further, in the prediction signal generating step in the video decoding method, the video decoding device includes a plurality of blocks including blocks separated by one block or more from the target block in the same frame as the block to be decoded. It is desirable to adaptively determine the probability distribution used when decoding the motion vector difference value based on the similarity of the motion vectors.

  A moving image encoding program according to the present invention causes a computer to function as the moving image encoding apparatus according to any one of the above-described aspects.

  A moving picture decoding program according to the present invention causes a computer to function as the moving picture decoding apparatus according to any one of the aspects described above.

  According to the present invention, encoding efficiency and accuracy can be improved. Also, the decoding efficiency and accuracy can be improved.

  A video encoding device and video decoding device according to an embodiment of the present invention will be described with reference to the drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and the overlapping description is abbreviate | omitted.

[First Embodiment]
A moving picture coding apparatus 10 according to the present invention will be described with reference to FIG. As shown in FIG. 1, the moving image encoding apparatus 10 includes a block divider 101, a frame memory 102, a motion detector 103, a prediction signal generator 104, and a motion vector memory 105 as functional components. A subtractor 106, a converter 107, a quantizer 108, an inverse quantizer 109, an inverse transformer 110, an adder 111, a motion predictor 112, a subtractor 113, and entropy coding. Device 114.

  An input video signal (moving image signal) input to the moving image encoding apparatus 10 is composed of a time series of image signals in units of frames (hereinafter referred to as “frame image signals”).

  The block divider 101 divides the frame image signal to be encoded of the input video signal input via the line L101 into a plurality of regions (blocks), and blocks the block to be encoded via the line L102 ( The image signal (hereinafter referred to as “target block”) is output. In this embodiment, the block is divided into 8 × 8 pixels, but may be divided into other sizes or shapes. The following encoding process is performed for each block.

  The frame memory 102 stores a previously encoded frame image signal (hereinafter referred to as “reference frame image signal”) input via the line L103, and outputs it via the line L104 as necessary.

  The motion detector 103 searches for an image signal pattern similar to the image signal pattern of the target block input via the line L102 within a predetermined range from the reference frame image signal input via the line L104. Then, a motion vector, which is a spatial displacement amount between both image signal patterns, is detected and output via a line L105. Further, the frame number referred to by the motion vector (hereinafter referred to as “reference frame number”) is detected and output via the line L115.

  The prediction signal generator 104 includes the motion vector signal value of the target block input via the line L105, the reference frame number input via the line L115, and the reference frame image input via the line L104. A predicted image signal for the target block is generated from the signal and output via the line L106.

  The motion vector memory 105 stores the motion vector input via the line L105 and the reference frame number input via the line L115 together with the motion vector and the reference frame number encoded in the past. Output via line L112 as necessary.

  The subtractor 106 subtracts the prediction image signal for the target block input via the line L106 from the image signal of the target block input via the line L102, generates a residual signal, Output via.

  The converter 107 orthogonally transforms the residual signal input via the line L107 and outputs a conversion coefficient via the line L108.

  The quantizer 108 quantizes the transform coefficient input via the line L108, and outputs the quantized transform coefficient via the line L109.

  The inverse quantizer 109 inversely quantizes the quantized transform coefficient input via the line L109, and outputs the restored transform coefficient via the line L110.

  The inverse transformer 110 performs inverse orthogonal transform on the transform coefficient input via the line L110, and outputs a residual signal restored via the line L111.

  The adder 111 adds the residual signal input via the line L111 and the prediction image signal for the target block input via the line L106, restores the image signal of the target block, and sets the line L103 to Output via.

  The motion predictor 112 uses the motion vector of the previously encoded block input via the line L112, the corresponding reference frame number, and the reference frame number of the target block input via the line L115. Then, the motion vector prediction value of the target block is calculated according to the flow described later, and is output via the line L113.

  The subtractor 113 subtracts the motion vector prediction value of the target block input via the line L113 from the motion vector value of the target block input via the line L105 to generate a motion vector difference value. Then, the motion vector difference value is output via the line L114.

  The entropy encoder 114 refers to the quantized transform coefficient input via the line L109, the motion vector difference value input via the line L114, and the target block input via the line L115. The frame number is encoded and multiplexed, and then the compressed stream is output to the line L116.

  Next, the motion predictor 112 will be described in detail with reference to FIG. FIG. 2 is a schematic diagram for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  In FIG. 2, the target block is H, and the frame F001 is a frame including the block H. The frame F002 is an already encoded frame, and the frame F001 is a frame adjacent in the time direction.

  In frame F001, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the upper leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F002, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block H. Block B is a block including a pixel immediately above the upper leftmost pixel of block A, and block C is a block including a pixel immediately left of the uppermost left pixel of block A.

  FIG. 3 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  First, the motion predictor 112 acquires the motion vectors of the encoded blocks A, B, C, E, and F input via the line L112 and the corresponding reference frame numbers (step S001).

  Note that the motion vector acquired in step S001 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

なお、このスケーリング処理は必須ではなく、スケーリング処理を行わない場合は各ブロックに対応する参照フレーム番号は必要ない。 Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by the following equation (1).

Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  Further, if there is no encoded block itself or a motion vector of the encoded block in step S001, it is considered that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, and the difference of the motion vector of block C and the motion vector of block F are calculated (step S002).

  Next, it is determined whether at least one of the two differences calculated in step S002 is equal to or less than a threshold value (step S003).

  As a result of the determination in step S003, if at least one of the two differences is equal to or less than the threshold value (step S003: YES), the process proceeds to step S004. If the two differences are both greater than the threshold (step S003: NO), the process proceeds to step S005.

  In step S004, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  In step S005, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F.

  Finally, the determined value PMVi (i = x, y) is output (step S006), and the process ends.

  Note that the determination in step S003 may determine whether both of the two differences are equal to or less than a threshold value.

  In step S005, the motion vector prediction value of block H is the same as the motion vector of block F, but the motion vector may be a zero vector (that is, no motion vector is predicted). An arbitrary motion vector determined in advance may be used as the predicted value.

  Further, the motion vector prediction value in the motion predictor 112 may be determined according to the flowchart of FIG.

  In FIG. 4, it is assumed that predicted values are independently determined for the x and y components of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 acquires reference frame numbers corresponding to motion vectors of encoded blocks A, B, C, D, E, and F input via the line L112 (step S101).

  Note that the motion vector acquired in step S101 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no coded block itself or a motion vector of the coded block in step S101, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And four different motion vector differences of the block F are calculated (step S102).

  Next, it is determined which combination of the four differences calculated in step S102 has the smallest difference. (Step S103).

  As a result of the determination in step S103, if the combination that gives the smallest difference is block B and block E, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. In the case of block C and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. In the case of block D and block E, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F. In the case of block D and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block E (step S104).

  Finally, the value PMVi (i = x, y) determined in step S104 is output (step S105), and the process ends.

  The series of processes shown in the flowchart of FIG. 4 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Further, the procedure for determining the motion vector prediction value in the motion predictor 112 may follow the flowchart described in FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 acquires reference frame numbers corresponding to motion vectors of encoded blocks A, B, C, D, E, and F input via the line L112 (step S201).

  Note that the motion vector acquired in step S201 is preferably scaled based on the temporal distance between the frame including the target block and the frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S201, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block D and the motion vector of block E, and the difference between the motion vector of block D and the motion vector of block F are calculated. If the former is larger (step S202: YES), the process is performed. The process proceeds to S203. If not (NO at step S202), the process proceeds to step S204 (step S202).

  In step S203, the difference between the motion vector of block E and the motion vector of block B, and the difference between the motion vector of block D and the motion vector of block F are calculated. If the former is larger (step S203: YES), the process is performed. Proceed to step S205. If not (NO at step S203), the process proceeds to step S206.

  In step S204, the difference between the motion vector of block C and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block A are calculated. If the former is larger (step S204: YES), the process is performed. The process proceeds to step S207. If not (NO at step S204), the process proceeds to step S208.

  In step S205, the motion vector prediction value of block H is determined to be the same value as the motion vector of block E.

  In step S206, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  In step S207, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F.

  In step S208, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  Finally, the determined motion vector prediction value PMVi (i = x, y) is output (step S209), and the process ends.

  The series of processes shown in the flowchart of FIG. 5 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 2 may be determined as follows. The procedure for determining the motion vector prediction value at that time may follow any of the flowcharts described in FIG. 3, FIG. 4 or FIG.

  The target block is H, and the frame F001 is a frame including the block H. Frame F002 is an already encoded frame.

  In frame F001, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the upper leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In frame F002, block A is a block that is referenced by the motion vector of block D, a block that is referenced by the motion vector of block E, or a block that is referenced by the motion vector of block F. Also in this case, the block B is a block including a pixel immediately above the uppermost leftmost pixel of the block A, and the block C is a block including a pixel rightward of the uppermost leftmost pixel of the block A.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block D or the block E refers to It is conceivable that the block or the block referenced by block F may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these blocks or the block with the highest ratio in the spanning area The motion vector is used as a motion vector of a block referred to by the corresponding block (D, E, or F).

  Further, the motion vector prediction value in the motion predictor 112 may be determined as follows.

  FIG. 6 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  In FIG. 6, the target block is H, and the frame F101 is a frame including the block H. A frame F102 is an already encoded frame, and is a frame adjacent to the frame F101 in the time direction. The frame F103 is a frame that has already been encoded, and the frame F102 is a frame that is adjacent in the time direction.

  In the frame F101, the block E is a block including the pixel immediately above the upper leftmost pixel of the block H, the block F is a block including the pixel immediately above the uppermost left pixel of the block H, and the block G is uppermost left of the block H. A block including a pixel immediately to the left of the pixel.

  In the frame F102, the block D is a block including a pixel at the same spatial position as the uppermost left pixel of the block H, the block B is a block including a pixel immediately above the uppermost left pixel of the block D, and the block C is the highest block of the block D. It is assumed that the block includes a pixel immediately to the left of the upper left pixel.

  In the frame F103, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block D.

  FIG. 7 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  In FIG. 7, it is assumed that predicted values are independently determined for the x and y components of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 acquires reference frame numbers corresponding to motion vectors of encoded blocks A, B, C, D, E, F, and G input via the line L112 (step S301). .

  Note that the motion vector acquired in step S301 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no coded block itself or a motion vector of the coded block in step S301, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S302).

  Next, it is determined which combination of the five differences calculated in step S302 has the smallest difference. (Step S303).

  As a result of the determination in step S303, when the combination that gives the smallest difference is the block B and the block F, the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block B. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C. In the case of block A and block D, the motion vector prediction value of block H is determined to be the same value as the motion vector of block D. In the case of block E and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block G. In the case of block E and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F. (Step S304).

  Finally, the value PMVi (i = x, y) determined in step S304 is output (step S305), and the process ends.

  The series of processes shown in the flowchart of FIG. 7 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Also, the difference calculated in step S302 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. At that time, the smallest combination is determined in step S303. In step S304, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. Determined by value. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C. In the case of block A and block D, the motion vector prediction value of block H is determined to be the same value as the motion vector of block D.

  Further, the difference calculated in step S302 may be two types, that is, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. At that time, the smallest combination is determined in step S303. In step S304, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. Determined by value. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C.

  Further, in order to increase the prediction accuracy when the region including the target block is moved greatly, the positional relationship of the blocks shown in FIG. 6 may be determined as follows.

  The target block is H, and the frame F101 is a frame including the block H. The frame F102 is an already encoded frame. A frame F103 is an already encoded frame.

  In the frame F101, the block E is a block including the pixel immediately above the upper leftmost pixel of the block H, the block F is a block including the pixel immediately above the uppermost left pixel of the block H, and the block G is uppermost left of the block H. A block including a pixel immediately to the left of the pixel.

  In the frame F102, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In the frame F103, the block A is a block referred to by the motion vector of the block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  In addition, the motion vector predictor 112 determines the motion vector prediction values of blocks that are spatially separated by one or more pixels as shown below, or blocks in a frame that is separated by one or more frames in the time direction. It may be performed based on a motion vector.

  Such a method for determining a motion vector prediction value has an effect of obtaining a motion vector prediction value with high accuracy by performing prediction reflecting a large number of motion vectors. It also has the effect of countering the effects of noise and local changes.

  In addition, the motion vector prediction value in the motion predictor 112 is a method for determining a plurality of motion vectors as described below, instead of using the motion vector of a block that has been encoded in the past as it is as a motion vector prediction value. One of them may be selected and determined according to the determination method.

  FIG. 8 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  In FIG. 8, the target block is J, and the frame F201 is a frame including the block J. The frame F202 is a frame that has already been encoded, and the frame F201 is a frame that is adjacent in the time direction. The frame F203 is a frame that has already been encoded, and the frame F202 is a frame that is adjacent in the time direction. The frame F204 is an already encoded frame, and the frame F203 is a frame adjacent in the time direction.

  In frame F201, block I is a block including the pixel immediately above the uppermost leftmost pixel of block J, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block J, a block F including the pixel immediately left of the uppermost leftmost pixel of the block J, a block E including a pixel immediately above the leftmost pixel of the block F, and a block Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E.

  In the frame F202, the block C is a block including a pixel at the same spatial position as the upper left pixel of the block J.

  In the frame F203, the block B is a block including a pixel at the same spatial position as the upper left pixel of the block C.

  In the frame F204, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block B.

  FIG. 9 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 112. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 receives the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, G, H, and I input via the line L112 and the line L115. The reference frame number of the target block input via is acquired (step S401).

  Note that the motion vector acquired in step S401 is preferably scaled based on the temporal distance between the frame including the target block and the frame referenced by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S401, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, variance of motion vector of block D, motion vector of block E and motion vector of block F, motion vector of block G and block Three sets of variances of H motion vector and block I motion vector variance are calculated. (Step S402)
Next, it is determined which combination of the three variances calculated in step S402 has the smallest variance. (Step S403).

  As a result of the determination in step S403, when the combination that gives the smallest difference is block A, block B, and block C, the motion vector prediction value of block J is the motion vector of block A, the motion vector of block B, and the block The average value of C motion vectors is determined. In the case of block D, block E, and block F, the motion vector prediction value of block J is determined as the average value of the motion vector of block D, the motion vector of block E, and the motion vector of block F. In the case of block G, block H, and block I, the motion vector prediction value of block J is determined as the average value of the motion vector of block G, the motion vector of block H, and the motion vector of block I (step S404). .

  Finally, the value PMVi (i = x, y) determined in step S404 is output (step S405), and the process ends.

  The series of processes shown in the flowchart of FIG. 9 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  In addition, although variance is used as a value representing the similarity of motion vectors in step S402, a standard deviation, a difference value of motion vectors between blocks, or an average of the plurality of difference values may be used.

  In step S404, the average value of the three motion vectors is set as the predicted value of the motion vector, but an intermediate value of the three motion vectors or any one of the values may be set as the predicted value.

  In FIG. 9, the motion vector prediction value is determined by referring to the motion vectors of three blocks continuous in the time direction and the spatial direction, but this may be only in the time direction. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 8 may be determined as follows.

  The target block is J, the frame F201 is a frame including the block J, and the frames F202, F203, and F204 are already encoded frames.

  In frame F201, block I is a block including the pixel immediately above the uppermost leftmost pixel of block J, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block J, a block F including the pixel immediately left of the uppermost leftmost pixel of the block J, a block E including a pixel immediately above the leftmost pixel of the block F, and a block Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E.

  In the frame F202, the block C is a block including a pixel referred to by the motion vector of the block I, a block referred to by the motion vector of the block F, or a block including the pixel immediately above the upper left pixel of the block J (block K)) is referred to by the motion vector.

  In frame F203, block B is a block that is referenced by the motion vector of block C.

  In frame F204, block A is a block that is referenced by the motion vector of block B.

  If the above method is used, block I, block F, block K, block C, block block when detecting the position of a similar image signal pattern with accuracy below the size of each block and detecting the motion vector It is possible that the block referenced by B spans multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the motion vector of the block with the highest ratio in the spanning area is applicable. This block is used as a motion vector of a block referred to by a block (I or F or K or C or B).

  Further, the motion vector prediction value in the motion predictor 112 is determined by the motion vector of a block spatially separated by one or more pixels as shown below, or a block in a frame separated by one or more frames in the time direction. It may be performed based on a motion vector.

  Such a method for determining a motion vector prediction value has an effect of obtaining a motion vector prediction value with high accuracy by performing prediction reflecting a large number of motion vectors. It also has the effect of countering the effects of noise and local changes.

  In addition, the motion vector prediction value in the motion predictor 112 is a method for determining a plurality of motion vectors as described below, instead of using the motion vector of a block that has been encoded in the past as it is as a motion vector prediction value. One of them may be selected and determined according to the determination method.

  FIG. 10 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  In FIG. 10, the target block is M, and the frame F301 is a frame including the block M. The frame F302 is a frame that has already been encoded, and the frame F301 is a frame that is adjacent in the time direction. The frame F303 is an already encoded frame, and is adjacent to the frame F302 in the time direction. The frame F304 is an already encoded frame, and the frame F303 is a frame adjacent in the time direction.

  In frame F301, block I is a block including the pixel immediately above the uppermost leftmost pixel of block M, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block M, a block F including the pixel immediately left of the leftmost pixel of the block M, and a block E including a pixel immediately left of the uppermost left pixel of the block F Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E. Also, block K is a block including the pixel immediately above the upper left pixel of block M, block L is a block including the pixel immediately above the upper left pixel of block K, and block J is the upper right pixel of block M. A block including a pixel on the right upper side of.

  In the frame F302, the block C is a block including a pixel at the same spatial position as the upper left pixel of the block M.

  In the frame F303, the block B is a block including a pixel at the same spatial position as the upper left pixel of the block C.

  In the frame F304, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block B.

  FIG. 11 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 112.

  First, the motion predictor 112 corresponds to the motion vectors of the encoded blocks A, B, C, D, E, F, G, H, I, J, K, and L that are input via the line L112. A reference frame number is acquired (step S501).

  Note that the motion vector acquired in step S501 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no coded block itself or a motion vector of the coded block in step S501, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of x components of block A motion vector, block B motion vector and block C motion vector, block D motion vector, block E motion vector and block F motion vector variance, block The three sets of x component variances of the J motion vector, the block K motion vector, and the block I motion vector variance are calculated. (Step S502)
Next, it is determined which combination of the three variances calculated in step S502 has the smallest variance. (Step S503).

  As a result of the determination in step S503, when the combination that gives the smallest difference is block A, block B, and block C, the x component of the motion vector prediction value of block M is the motion vector of block A and the motion of block B. The average value of the x component of the vector and the motion vector of block C is determined. In the case of block D, block E, and block F, the x component of the motion vector prediction value of block M is determined as the average value of the motion vector of block D, the motion vector of block E, and the x component of the motion vector of block F Is done. In the case of block J, block K, and block I, the x component of the motion vector prediction value of block M is determined as the average value of the motion vector of block J, the motion vector of block K, and the x component of the motion vector of block I (Step S504).

Subsequently, variance of y component of motion vector of block A, motion vector of block B and motion vector of block C, variance of y component of motion vector of block G, motion vector of block H and motion vector of block I, block Three sets of variances of the y component of the L motion vector, the motion vector of the block K, and the variance of the y component of the motion vector of the block F are calculated. (Step S505)
Next, it is determined which combination of the three variances calculated in step S505 has the smallest variance. (Step S506).

  As a result of the determination in step S506, when the combination that gives the smallest difference is block A, block B, and block C, the y component of the motion vector prediction value of block M is the motion vector of block A and the motion of block B. The average value of the y component of the vector and the motion vector of block C is determined. In the case of block G, block H, and block I, the y component of the motion vector prediction value of block M is determined as the average value of the y component of the motion vector of block G, the motion vector of block H, and the motion vector of block I Is done. In the case of block L, block K, and block F, the y component of the motion vector prediction value of block M is determined as the average value of the motion vector of block L, the motion vector of block K, and the y component of the motion vector of block F (Step S507).

  Finally, the value PMVi (i = x, y) determined in step S507 is output (step S508), and the process ends.

  Further, although variance is used as a value representing the similarity of motion vectors in step S502 and step S505, a standard deviation, a difference value of motion vectors between blocks, or an average of the plurality of difference values may be used. .

  In step S504 and step S507, the average value of the three motion vectors is used as the predicted value of the motion vector, but an intermediate value of the three motion vectors or any one of the values may be used as the predicted value.

  In FIG. 11, the motion vector prediction value is determined by referring to the motion vectors of three blocks continuous in the time direction and the space direction, but only the space direction may be used. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 10 may be determined as follows.

  The target block is M, the frame F301 is a frame including the block M, and the frames F302, F303, and F304 are already encoded frames.

  In frame F301, block I is a block including the pixel immediately above the uppermost leftmost pixel of block M, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block M, a block F including the pixel immediately left of the leftmost pixel of the block M, and a block E including a pixel immediately left of the uppermost left pixel of the block F Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E. Also, block K is a block including the pixel immediately above the upper left pixel of block M, block L is a block including the pixel immediately above the upper left pixel of block K, and block J is the upper right pixel of block M. A block including a pixel on the right upper side of.

  In the frame F302, the block C is a block that is referenced by the motion vector of the block I, a block that is referenced by the motion vector of the block K, or a block that is referenced by the motion vector of the block F.

  In the frame F303, the block B is a block referred to by the motion vector of the block C.

  In frame F304, block A is a block that is referenced by the motion vector of block B.

  If the above method is used, block I, block K, block F, block C, block block when detecting the position of a similar image signal pattern with accuracy below the size of each block and detecting the motion vector It is possible that the block referenced by B spans multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the motion vector of the block with the highest ratio in the spanning area is applicable. This block is used as a motion vector of a block to be referenced by a block (I or K or F or C or B).

  By the way, when calculating a motion vector prediction value, a peripheral motion vector is effectively used by weighting as follows, instead of selecting a part of motion vectors from previously encoded motion vectors in the vicinity. A method may be adopted. In this case, by effectively using surrounding motion vectors by weighting, it is possible to reduce the influence of local motion vector changes, obtain a predicted value reflecting the majority of motion vector changes, and improve prediction accuracy. be able to.

  In the following, the peripheral motion vector utilization method by weighting is applied to (a) an example applied to the block of FIG. 2, (b) an example applied to the block of FIG. 6, and (c) an example applied to the block of FIG. These will be described in order.

  First, (a) as an example applied to the block of FIG. 2, for example, the flowchart of FIG. That is, the procedure for determining the motion vector prediction value in the motion predictor 112 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 acquires reference frame numbers corresponding to motion vectors of encoded blocks A, B, C, D, E, and F input via the line L112 (step S3001).

  Note that the motion vector acquired in step S3001 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S3001, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block E, the difference MViD — 1 (i = x, y) between the motion vector of block C and the motion vector of block F, A difference MViD_2 (i = x, y) between the motion vector and the motion vector of the block E and a difference MViD_3 (i = x, y) between the motion vector of the block D and the motion vector of the block F are calculated (step S3002).

最後に、算出された動きベクトル予測値PMVi(i=x,y)を出力し（ステップＳ３００４）、処理を終了する。 Next, based on the following equation (2) using the function F having each difference as a variable calculated in step S3002, all the motion vectors of the blocks A, F, and E are effectively used, and a motion vector prediction value is obtained. PMVi (i = x, y) is calculated (step S3003).

Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3004), and the process ends.

  The series of processes shown in the flowchart of FIG. 57 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Moreover, as the function F used by Formula (2), you may employ | adopt the function which calculates the reciprocal number of a variable (difference), for example. In this case, for example, when the difference MViD — 0 (i = x, y) between the motion vector of the block B and the motion vector of the block E is small, the reciprocal of the difference, that is, a large value becomes the weighting coefficient of the motion vector of the block A. The weight of the motion vector of block A can be increased.

  Next, (b) as an example applied to the block of FIG. 6, for example, the flowchart of FIG. That is, the procedure for determining the motion vector prediction value in the motion predictor 112 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 acquires reference frame numbers corresponding to motion vectors of encoded blocks A, B, C, D, E, F, and G input via the line L112 (step S3101). .

  Note that the motion vector acquired in step S3101 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S 3101, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block F, the difference MViD — 1 (i = x, y) between the motion vector of block C and the motion vector of block G, The difference MViD — 2 (i = x, y) between the motion vector and the motion vector of block D, the difference MViD — 3 (i = x, y) between the motion vector of block E and the motion vector of block F, the motion vector of block E and the block G Five types of motion vector differences MViD — 4 (i = x, y) are calculated (step S3102).

最後に、算出された動きベクトル予測値PMVi(i=x,y)を出力し（ステップＳ３１０４）、処理を終了する。 Next, all the motion vectors of the blocks B, C, D, G, and F are effectively used based on the following formula (3) using the function F having each difference as a variable calculated in step S3102. A motion vector prediction value PMVi (i = x, y) is calculated (step S3103).

Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3104), and the process ends.

  The series of processes shown in the flowchart of FIG. 58 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be. Moreover, as the function F used by Formula (3), you may employ | adopt the function which calculates the reciprocal number of a variable (difference), for example. In this case, for example, when the difference MViD — 0 (i = x, y) between the motion vector of the block B and the motion vector of the block F is small, the reciprocal of the difference, that is, a large value becomes the weighting coefficient of the motion vector of the block B. The weight of the motion vector of block B can be increased.

  Further, the difference calculated in step S3102 is the difference MViD_0 (i = x, y) between the motion vector of block B and the motion vector of block F, and the difference MViD_1 (i = x) of the motion vector of block C and the motion vector of block G. , y). At this time, in step S3103, a function F (MV iD — 2 (i = x, y)), a function F (MV iD — 3 (i = x, y)), and a function F (MV iD — 4 (i = x, y)) are set. The motion vector prediction value PMVi (i = x, y) may be calculated with 0). In this case, the motion vector prediction value PMVi (i = x, y) can be calculated based on only the motion vector difference in the temporal direction without including the motion vector difference in the spatial direction.

  Further, the difference calculated in step S3102 is the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block F, and the difference MViD — 1 (i = x) of the motion vector of block C and the motion vector of block G. , y), the difference MViD_2 (i = x, y) between the motion vector of the block A and the motion vector of the block D may be used. At this time, in step S3103, the function F (MV iD — 3 (i = x, y)) and the function F (MV iD — 4 (i = x, y)) using the difference not calculated as variables are set to 0 and the motion vector prediction value PMVi (i = x, y) may be calculated. In this case, the motion vector prediction value PMVi (i = x, y) can be calculated based only on the temporal motion vector difference including the motion vector difference between the blocks one frame or more away from the target block. it can.

  Next, (c) as an example applied to the block of FIG. 8, there is a flowchart of FIG. 59, for example. That is, the procedure for determining the motion vector prediction value in the motion predictor 112 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 112 receives the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, G, H, and I input via the line L112 and the line L115. The reference frame number of the target block input via is acquired (step S3201).

  Note that the motion vector acquired in step S3201 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S3201, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

最後に、算出された動きベクトル予測値PMVi(i=x,y)を出力し（ステップＳ３２０４）、処理を終了する。 Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, MViV_0 (i = x, y), variance of motion vector of block D, motion vector of block E and motion vector of block F Three sets of variances are calculated: MViV_1 (i = x, y), block G motion vector, block H motion vector, and block I motion vector variance MViV_2 (i = x, y). (Step S3202)
Next, based on the following equation (4) using the function F having each variance as a variable calculated in step S3202, all nine motion vectors of the blocks A to I are effectively used, and a motion vector prediction value is obtained. PMVi (i = x, y) is calculated (step S3203).

Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3204), and the process ends.

  The series of processes shown in the flowchart of FIG. 59 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be. Moreover, as the function F used by Formula (4), you may employ | adopt the function which calculates the reciprocal number of a variable (dispersion), for example. In this case, for example, when the variance MViV — 0 (i = x, y) of the motion vector of block A, the motion vector of block B, and the motion vector of block C is small, the reciprocal of the variance, that is, a large value is the motion vector of block A Since the weighting coefficient of the average value of the motion vector of block B and the motion vector of block C can be increased, the weight of the average value of the motion vector of block A, the motion vector of block B, and the motion vector of block C can be increased. .

  Further, although variance is used as a value representing the similarity of motion vectors in step S3202, standard deviation, a difference value of motion vectors between blocks, or an average of the plurality of difference values may be used.

  In step S3203, the average value of the three motion vectors is set as the predicted value of the motion vector, but an intermediate value of the three motion vectors or any one of the values may be set as the predicted value.

  In FIG. 59, the motion vector prediction value is determined with reference to three blocks of motion vectors that are continuous in the time direction and the spatial direction, but this may be only in the time direction. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  Further, the variance calculated in step S3202 is the variance MViV_1 (i = x, y) of the motion vector of block D, the motion vector of block E, and the motion vector of block F, the motion vector of block G, and the motion vector of block H. Only two sets of block vector motion vector variance MViV — 2 (i = x, y) may be used. At this time, in step S3203, the motion vector prediction value PMVi (i = x, y) may be calculated by setting the function F (MViV — 0 (i = x, y)) with the variance not calculated as a variable to zero. In this case, the motion vector prediction value PMVi (i = x, y) can be calculated only from the spatial direction including a plurality of blocks separated by one block or more from the target block.

  Finally, a moving picture coding program that causes a computer to operate as the moving picture coding apparatus 10 will be described. FIG. 51 shows a configuration of a moving image encoding program P10 according to the present embodiment, and this moving image encoding program P10 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. FIG. 63 is a diagram showing a hardware configuration of a computer for executing a program stored in the recording medium, and FIG. 64 is a perspective view of the computer for executing the program stored in the recording medium.

  As shown in FIG. 63, the computer 80 includes a reading device 62 such as a flexible disk drive device, a CD-ROM drive device, a DVD drive device, a working memory (RAM) 64, a program stored in the recording medium 60, and the like. A memory 66, a display 68, a mouse 70, a keyboard 72, a communication device 74 that transmits and receives data and the like, and a CPU 76 that controls execution of a program. When the recording medium 60 is inserted into the reading device 62, the computer 80 can access the moving image encoding program P10 stored in the recording medium 60 from the reading device 62, and the moving image encoding program P10 allows the moving image encoding program P10 to It becomes possible to operate as the image encoding device 10.

  As shown in FIG. 64, the moving image encoding program P10 may be provided via a network as a computer data signal 90 superimposed on a carrier wave. In this case, the computer 80 can store the moving image encoding program P10 received by the communication device 74 in the memory 66 and execute the moving image encoding program P10.

  As shown in FIG. 51, the moving picture encoding program P10 includes a main module P101 that supervises processing, a block division module P102, a transform module P103, a quantization module P104, an inverse quantization module P105, and an inverse transform. A module P106, a storage module P107, a prediction signal generation module P108, a motion prediction module P109, a motion detection module P110, an addition module P111, a subtraction module P112, and an entropy encoding module P113 are provided.

  Block division module P102, transform module P103, quantization module P104, inverse quantization module P105, inverse transform module P106, prediction signal generation module P108, motion prediction module P109, motion detection module P110, addition module P111, entropy encoding module P113 The functions realized by the computer 80 are the block divider 101, transformer 107, quantizer 108, inverse quantizer 109, inverse transformer 110, prediction signal generator 104, motion predictor 112, motion of FIG. The functions of the detector 103, the adder 111, and the entropy encoder 114 are the same. The functions that the storage module P107 realizes in the computer 80 are the same as the functions of the frame memory 102 and the motion vector memory 105. The function that the subtraction module P112 realizes in the computer 80 is the same as the function of the subtractors 106 and 113.

[Second Embodiment]
The moving picture coding apparatus 20 according to the present invention will be described with reference to FIG. As shown in FIG. 12, the moving image coding apparatus 20 includes a block divider 201, a frame memory 202, a motion detector 203, a prediction signal generator 204, and a motion vector memory 205 as functional components. A subtractor 206, a converter 207, a quantizer 208, an inverse quantizer 209, an inverse transformer 210, an adder 211, a motion predictor 212, a probability switching signal generator 213, And an entropy encoder 114.

  An input video signal (moving image signal) input to the moving image encoding device 20 is composed of a time series of image signals in units of frames (hereinafter referred to as frame image signals).

  The block divider 201 divides the frame image signal to be encoded of the input video signal input via the line L201 into a plurality of areas (blocks), and blocks the block to be encoded via the line L202 ( Hereinafter, the image signal of the target block) is output. In this embodiment, the block is divided into 8 × 8 pixels, but may be divided into other sizes or shapes. The following encoding process is performed for each block.

  The frame memory 202 stores a previously encoded frame image signal (hereinafter referred to as a reference frame image signal) input via the line L203 and outputs it via the line L204 as necessary.

  The motion detector 203 searches an image signal pattern similar to the image signal pattern of the target block input via the line L202 within a predetermined range from the reference frame image signal input via the line L204. Then, a motion vector, which is a spatial displacement amount between both image signal patterns, is detected and output via the line L205. Also, the frame number (hereinafter referred to as the reference frame number) referred to by the motion vector is detected and output via the line L216.

  The prediction signal generator 204 includes the motion vector signal value of the target block input via the line L205, the reference frame number input via the line L216, and the reference frame image input via the line L204. A predicted image signal for the target block is generated from the signal and output via the line L206.

  The motion vector memory 205 stores a motion vector input via the line L205 and a reference frame number input via the line L215, together with a previously encoded motion vector and reference frame number. Output via line L212 as necessary.

  The subtracter 206 subtracts the prediction image signal for the target block input via the line L206 from the image signal of the target block input via the line L202, generates a residual signal, Output via.

  The converter 207 performs orthogonal transform on the residual signal input via the line L207 and outputs a conversion coefficient via the line L208.

  The quantizer 208 quantizes the transform coefficient input via the line L208 and outputs the quantized transform coefficient via the line L209.

  The inverse quantizer 209 inversely quantizes the quantized transform coefficient input via the line L209 and outputs the restored transform coefficient via the line L210.

  The inverse transformer 210 performs an inverse orthogonal transform on the transform coefficient input via the line L210 and outputs a restored residual signal via the line L211.

  The adder 211 adds the residual signal input via the line L211 and the prediction image signal for the target block input via the line L206, restores the image signal of the target block, and sets the line L203 Output via.

  The motion predictor 212 uses the motion vector of the previously encoded block input via the line L212, the corresponding reference frame number, and the reference frame number of the target block input via the line L216. In accordance with the flow to be described later, the index indicating the motion vector prediction value of the target block is determined, the motion vector difference value that is the difference between the motion vector prediction value and the actual motion vector is calculated, and the motion vector prediction value is obtained via the line L213. The index indicating is output. Also, the motion vector difference value is output via the line L214.

  The probability switching signal generator 213 inputs the motion vector of the previously encoded block input via the line L212, the corresponding reference frame number, and the reference frame of the target block input via the line L216. From the number, a probability switching signal indicating a probability table to be switched according to a flow to be described later is generated and output via a line L215.

  The entropy encoder 214 receives the quantized transform coefficient input via the line L209, the motion vector difference value input via the line L214, and the target block input via the line L216. The reference frame number is encoded. In addition, based on the probability switching signal input via the line L215, the probability table used when performing entropy coding is switched by the method described later, and then the motion vector prediction value input via the line L213 The index indicating is encoded. The encoded transform coefficient, motion vector difference value, and index indicating the motion vector prediction value are multiplexed, and then the compressed stream is output to line L216.

  Switching the probability table used when entropy encoding is performed as described above is performed by encoding each index indicating a motion vector prediction value with an appropriate code amount based on the appearance frequency, thereby increasing the index encoding efficiency. There is an effect to increase.

  Next, the motion predictor 212 will be described in detail with reference to FIG. FIG. 13 is a schematic diagram for explaining a procedure for determining an index indicating a motion vector prediction value in the motion predictor 112.

  In FIG. 13, the target block is H, and the frame F401 is a frame including the block H. The frame F402 is an already encoded frame, and is adjacent to the frame F401 in the time direction.

  In frame F401, block D is a block including the pixel immediately above the upper left pixel of block H, block E is a block including the pixel immediately above the upper left pixel of block H, and block F is the upper left pixel of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F402, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block H. Block B is a block including a pixel immediately above the upper leftmost pixel of block A, and block C is a block including a pixel immediately left of the uppermost left pixel of block A.

  FIG. 14 is a flowchart for explaining a procedure for determining an index indicating a motion vector prediction value in the motion predictor 212. Assume that the indices of the x and y components of the motion vector are determined independently. Specifically, an index indicating the predicted value of the x component of the motion vector is first determined and output, and then an index indicating the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 212 receives the motion vectors of the encoded blocks A, E, and F input via the line L212, the corresponding reference frame numbers, and the target block input via the line L205. The motion vector of H and the reference frame number of the target block input via the line L216 are acquired (step S601).

  Note that the motion vector acquired in step S601 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and E obtained by scaling based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referenced by the motion vector of the target block , The magnitude of the motion vector of F can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S601, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block A and the motion vector of block H, the difference between the motion vector of block E and the motion vector of block H, and the difference between the motion vector of block F and the motion vector of block H Is calculated (step S602).

  Next, it is determined which combination of the three difference values calculated in step S602 has the smallest difference value. (Step S603).

  If the combination that gives the smallest difference is the block A and the block H as a result of the determination in step S603, the index Index_i = indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block A 0 (i = x, y) is generated, and the difference between the motion vectors of block A and block H is set as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block E and the block H, the index Index_i = 1 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block E And the difference between the motion vectors of the block E and the block H is a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block F and the block H, the index Index_i = 2 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block F The difference between the motion vectors of block A and block H is set as a motion vector difference value MViD (i = x, y) (step S604).

  Subsequently, the index Index_i (i = x, y) determined in step S604 is output (step S605).

  Finally, the motion vector difference value MViD (i = x, y) determined in step S604 is output (step S606), and the process ends.

  In step S602, the motion vector prediction value is determined using the difference between the motion vector prediction value and the actual motion vector. Instead, encoding is actually performed to optimize the code amount and distortion (RD). The motion vector prediction value may be determined by optimization).

  The series of processes shown in the flowchart of FIG. 14 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into an index indicating the x component and y component of the motion vector. The motion vector difference value may be determined simultaneously.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 13 may be determined as follows.

  The target block is H, and the frame F401 is a frame including the block H. A frame F402 is an already encoded frame.

  In frame F401, block D is a block including the pixel immediately above the upper left pixel of block H, block E is a block including the pixel immediately above the upper left pixel of block H, and block F is the upper left pixel of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F402, the block A is a block that is referenced by the motion vector of the block D, a block that is referenced by the motion vector of the block E, or a block that is referenced by the motion vector of the block F. Also in this case, the block B is a block including a pixel immediately above the uppermost leftmost pixel of the block A, and the block C is a block including a pixel rightward of the uppermost leftmost pixel of the block A.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block D or the block E refers to It is conceivable that the block or the block referenced by block F may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these blocks or the block with the highest ratio in the spanning area The motion vector is used as a motion vector of a block referred to by the corresponding block (D, E, or F).

  In the embodiment described above, motion vector prediction is performed by combining a motion vector difference value obtained between blocks in the same frame and a motion vector difference value obtained between blocks in different frames. The motion vector may be predicted using only the motion vector difference value obtained between the blocks. In that case, among the combinations of the difference between the motion vectors of the blocks A and H, the difference between the motion vectors of the blocks B and H, and the difference between the motion vectors of the blocks C and H, the one having the smallest difference value may be used. Furthermore, the difference between the motion vector average value of blocks A, B, and C and the motion vector of block H may be added to the candidates.

  At this time, the details of the probability switching signal generator 213 preferably follow the flowchart of FIG. The probability switching signal indicates the likelihood of the occurrence probability of the index indicating the motion vector prediction value, and by switching the probability table based on this signal and performing entropy coding, the probability switching signal Indexes can be encoded with a large code amount.

  In the probability switching signal generator 213, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 213 receives the reference frame numbers corresponding to the motion vectors of the encoded blocks A, B, C, D, E, and F input via the line L212 and the line L216. The reference frame number of the target block input in step S701 is acquired.

  Note that the motion vector acquired in step S701 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  Also, if there is no encoded block itself or a motion vector of the encoded block in step S701, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And four different motion vector differences of block F are calculated (step S702).

  Next, it is determined which combination of the four differences calculated in step S702 has the smallest difference. (Step S703).

  As a result of the determination in step S703, if the combination that gives the smallest difference is block B and block E, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. A signal Signal_i (i = x, y) = 0 indicating the difference between the motion vectors of the block B and the block E is used as a difference MViD (i = x, y) at that time. When the combination that gives the smallest difference is the block C and the block F, the signal Signal_i (i = i = i) indicates that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block A. x, y) = 0 is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block F. When the combination that gives the smallest difference is the block D and the block E, the signal Signal_i (i = i = i) indicates that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block F. x, y) = 1 is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block E. When the combination that gives the smallest difference is the block D and the block F, the signal Signal_i (i = i = i) indicates that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block E. x, y) = 2 is generated, and the motion vector difference value MViD (i = x, y) at that time is set as the difference value between the motion vectors of the block D and the block F. (Step S704).

Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S704 are output (step S705), and the process ends.
The occurrence probability signal may be only Signal_i (i = x, y).

  The series of processes shown in the flowchart of FIG. 15 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the entropy encoder 214 switches a table of occurrence probability distribution of each index indicating a plurality of motion vector prediction values possessed in advance based on a probability switching signal input via the line L215. Specifically, the index with the highest probability of occurrence is calculated by Signal_i (i = x, y), the probability of occurrence is predicted by MViD (i = x, y), and switching to the probability table is performed accordingly.

  Further, the motion predictor 212 may be described below. The motion predictor 212 will be described in detail with reference to FIG. FIG. 16 is a schematic diagram for explaining a procedure for determining an index indicating a motion vector prediction value in the motion predictor 112.

  In FIG. 16, the target block is H, and the frame F501 is a frame including the block H. A frame F502 is an already encoded frame, and is a frame adjacent to the frame F501 in the time direction. A frame F503 is an already encoded frame, and is adjacent to the frame F502 in the time direction.

  In frame F501, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F502, the block D is a block including a pixel at the same spatial position as the upper leftmost pixel of the block H, the block B is a block including a pixel immediately above the uppermost left pixel of the block D, and the block C is the highest block of the block D. It is assumed that the block includes a pixel immediately to the left of the upper left pixel.

  In the frame F503, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block D.

  FIG. 17 is a flowchart for explaining a procedure for determining an index indicating a motion vector prediction value in the motion predictor 212. Assume that the indices of the x and y components of the motion vector are determined independently. Specifically, an index indicating the predicted value of the x component of the motion vector is first determined and output, and then an index indicating the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 212 obtains the motion vectors of the encoded blocks A, B, C, D, E, F, and G input via the line L212, the corresponding reference frame numbers, and the line L205. The motion vector of the target block H that is input via and the reference frame number of the target block input via the line L216 are acquired (step S801).

  Note that the motion vector acquired in step S801 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referenced by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S801, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block H, the difference of the motion vector of block C and the motion vector of block H, the difference of the motion vector of block D and the motion vector of block H, the motion vector of block G And the difference between the motion vector of block H and the difference between the motion vector of block F and the motion vector of block H are calculated (step S802).

  Next, it is determined which combination of the five difference values calculated in step S802 has the smallest difference value. (Step S803).

  If the combination that gives the smallest difference is the block B and the block H as a result of the determination in step S803, the index Index_i = indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block B 0 (i = x, y) is generated, and the difference between the motion vectors of block B and block H is set as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block C and the block H, an index Index_i = 1 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block C And the difference between the motion vectors of the block C and the block H is a motion vector difference value MViD (i = x, y). When the combination giving the smallest difference is the block D and the block H, the index Index_i = 2 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block D And the difference between the motion vectors of block D and block H is defined as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block G and the block H, an index Index_i = 3 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block G And the difference between the motion vectors of the block G and the block H is defined as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block F and the block H, the index Index_i = 4 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block F The difference between the motion vectors of the block F and the block H is set as a motion vector difference value MViD (i = x, y) (step S804).

  Subsequently, the index Index_i (i = x, y) determined in step S804 is output (step S805).

  Finally, the motion vector difference value MViD (i = x, y) determined in step S804 is output (step S806), and the process ends.

  In step S802, the motion vector prediction value is determined using the difference between the motion vector prediction value and the actual motion vector. Instead, encoding is actually performed to optimize the code amount and distortion (RD). The motion vector prediction value may be determined by optimization).

  Note that the series of processes shown in the flowchart of FIG. 17 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into an index indicating the x component and y component of the motion vector. The motion vector difference value may be determined simultaneously.

  Also, the difference calculated in step S802 is the difference between the motion vector of block B and the motion vector of block H, the difference between the motion vector of block C and the motion vector of block H, the motion vector of block D and the motion vector of block H. Three types of differences may be used. At that time, the smallest combination is determined in step S803. In step S804, when the combination that gives the smallest difference is the block B and the block H, the motion vector prediction value of the block H is the motion vector of the block B. An index Index_i = 0 (i = x, y) indicating the same value is generated, and a difference between motion vectors of block B and block H is set as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block C and the block H, an index Index_i = 1 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block C And the difference between the motion vectors of the block C and the block H is a motion vector difference value MViD (i = x, y). When the combination giving the smallest difference is the block D and the block H, the index Index_i = 2 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block D And the difference between the motion vectors of block D and block H is defined as a motion vector difference value MViD (i = x, y).

  Further, the difference calculated in step S302 may be two types, that is, the difference between the motion vector of block B and the motion vector of block H, and the difference between the motion vector of block C and the motion vector of block H. At that time, the smallest combination is determined in step S803. In step S804, when the combination that gives the smallest difference is the block B and the block H, the motion vector prediction value of the block H is the motion vector of the block B. An index Index_i = 0 (i = x, y) indicating the same value is generated, and a difference between motion vectors of block B and block H is set as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block C and the block H, an index Index_i = 1 (i = x, y) indicating that the motion vector prediction value of the block H is the same value as the motion vector of the block C And the difference between the motion vectors of the block C and the block H is a motion vector difference value MViD (i = x, y).

  Further, in order to increase the prediction accuracy when the region including the target block is moved greatly, the positional relationship of the blocks shown in FIG. 16 may be determined as follows.

  The target block is H, and the frame F501 is a frame including the block H. A frame F502 is an already encoded frame. A frame F503 is an already encoded frame.

  In frame F501, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F502, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In the frame F503, the block A is a block referred to by the motion vector of the block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  At this time, the details of the probability switching signal generator 213 preferably follow the flowchart of FIG. The probability switching signal indicates the likelihood of the occurrence probability of the index indicating the motion vector prediction value, and by switching the probability table based on this signal and performing entropy coding, the probability switching signal Indexes can be encoded with a large code amount.

  In the probability switching signal generator 213, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 213 sets the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, and G input via the line L212, and the line L216. The reference frame number of the target block that is input via is acquired (step S901).

  Note that the motion vector acquired in step S901 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S901, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S902).

  Next, it is determined which combination of the five differences calculated in step S902 has the smallest difference. (Step S903).

  As a result of the determination in step S903, if the combination that gives the smallest difference is block B and block F, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block B. A signal Signal_i = 0 (i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as a difference value between the motion vectors of the block B and the block F. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D. When the combination that gives the smallest difference is the block E and the block F, the signal Signal_i = 3 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block G. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block E and block F. When the combination that gives the smallest difference is the block E and the block G, the signal Signal_i = 4 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is used as the difference value between the motion vectors of block E and block G. (Step S904).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S904 are output (step S705), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y).

  The series of processes shown in the flowchart of FIG. 18 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  The difference calculated in step S902 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. Even in this case, the smallest combination is determined in step S903. In step S904, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D.

  Also, the difference calculated in step S902 may be two types, that is, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. Even in this case, the smallest combination is determined in step S903. In step S904, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G.

  Further, in order to increase the prediction accuracy when the region including the target block is moved greatly, the positional relationship of the blocks shown in FIG. 16 may be determined as follows.

  The target block is H, and the frame F501 is a frame including the block H. A frame F502 is an already encoded frame. A frame F503 is an already encoded frame.

  In frame F501, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F502, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In the frame F503, the block A is a block referred to by the motion vector of the block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  At this time, the entropy encoder 214 switches the occurrence probability distribution table of each index indicating the motion vector prediction values possessed in advance based on the probability switching signal input via the line L215. Specifically, the index with the highest probability of occurrence is calculated by Signal_i (i = x, y), the probability of occurrence is predicted by MViD (i = x, y), and switching to the probability table is performed accordingly.

  Instead of generating an index indicating a motion vector prediction value in the motion predictor 212, one of a plurality of motion vector prediction value determination methods may be selected and an index indicating the selected determination method may be generated. Details of this method will be described with reference to FIGS.

  FIG. 19 is a schematic diagram for explaining a procedure for selecting one of a plurality of motion vector prediction value determination methods in the motion predictor 212 and generating an index indicating the selected determination method.

  In FIG. 19, the target block is N, and the frame F601 is a frame including the block N. The frame F602 is a frame that has already been encoded, and is adjacent to the frame F601 in the time direction.

  In frame F601, block J is a block including the pixel immediately above the upper left pixel of block N, block K is a block including the pixel immediately above the upper left pixel of block N, and block L is the upper right pixel of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In the frame F602, the block E is a block including a pixel at the same spatial position as the upper left pixel of the block N. Block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, block C is the true of the upper right pixel of block E A block including the upper right pixel, a block D including the pixel immediately left of the lower leftmost pixel of the block E, a block F including a pixel right of the upper rightmost pixel of the block E, and a block G including the block E A block including the pixel directly below the leftmost pixel of block E, a block including the pixel immediately below the pixel below rightmost of block E, and a block I including the pixel directly below the rightmost pixel of block E A block containing

  FIG. 20 is a flowchart for explaining a procedure for selecting one of a plurality of motion vector prediction value determination methods in the motion predictor 212 and generating an index indicating the selected determination method.

  It is assumed that a prediction value determination method is independently selected for each of the motion vector x component and y component. Specifically, first, a predicted value determination method for the x component of the motion vector is selected and output, and then a predicted value determination method for the y component is selected and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 212 receives the motion vectors of the encoded blocks A, B, C, D, E, F, G, H, I, J, K, L, and M input via the line L212. The corresponding reference frame number, the motion vector of the target block N acquired via the line L205, and the reference frame number of the target block input via the line L216 are acquired (step S1001).

  Note that the motion vector acquired in step S1001 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no encoded block itself or a motion vector of the encoded block in step S1001, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the average value of the motion vectors of the blocks A, B, C, D, E, F, G, H, and I and the motion vector of the target block N, the motion vectors of the blocks B, D, E, F, and H Three differences are calculated: the difference between the average value and the motion vector of the target block N, the average value of the motion vectors of the blocks E, J, K, L, and M and the difference of the motion vector of the target block N (step S1002). .

  Subsequently, the combination having the smallest value among the three differences calculated in step S1002 is determined. (Step S1003).

  As a result of the determination in step S1003, when the combination that gives the smallest difference is the block A, B, C, D, E, F, G, H, I and the target block N, the motion vector prediction value of the block N is An index Index_i = 0 (i = x, y) indicating a method of determining an average value of motion vectors of blocks A, B, C, D, E, F, G, H, and I is generated, and the method is The difference between the motion vector prediction value when used and the motion vector of the target block is defined as a motion vector difference value MViD (i = x, y). When the combination that gives the smallest difference is the block B, D, E, F, H and the target block N, the motion vector predicted value of the block N is the average value of the motion vectors of the blocks B, D, E, F, H An index Index_i = 1 (i = x, y) indicating the method of determining the motion vector is generated, and the difference between the motion vector prediction value and the motion vector of the target block when using this method is calculated as a motion vector difference value MViD ( i = x, y). When the combination that gives the smallest difference average value is the block E, J, K, L, M and the target block N, the motion vector predicted value of the block N is the motion vector of the blocks E, J, K, L, M An index Index_i = 2 (i = x, y) indicating a method of determining an average value is generated, and a difference between a motion vector prediction value and a target block motion vector when using this method is calculated as a motion vector difference value MViD. (i = x, y) (step S1004).

  Next, the index Index_i (i = x, y) determined in step S1004 is output (step S1005).

  Finally, the motion vector difference value MViD (i = x, y) determined in step S1004 is output (step S1006), and the process ends.

  In step S1002, an intermediate value of the motion vectors of each block group may be used instead of the average value of the motion vectors of each block group. In this case, the index generated in step S1004 is an index indicating that an intermediate value of each block group is used as a motion vector prediction value instead of indicating that an average value of each block group is used as a motion vector prediction value. .

  In step S1002, the method of determining the motion vector prediction value is selected using the difference between the motion vector prediction value and the actual motion vector. Instead, encoding is actually performed to optimize the code amount and distortion ( A method for determining a motion vector prediction value may be selected by (RD optimization).

  In the embodiment shown in the flowchart of FIG. 20, one of the three motion vector prediction value determination methods is selected. However, any one of the three methods may be selected.

  The series of processes shown in the flowchart of FIG. 20 calculates a vector difference (specifically, a vector norm, etc.) without dividing the x component and the y component, and an index indicating how to determine each of the x component and the y component. May be selected simultaneously.

  Furthermore, in order to improve the prediction accuracy when the region including the target block is moved greatly, the positional relationship of the blocks shown in FIG. 19 may be determined as follows.

  The target block is N, and the frame F601 is a frame including the block N. Frame F602 is an already encoded frame.

  In frame F601, block J is a block including the pixel immediately above the upper left pixel of block N, block K is a block including the pixel immediately above the upper left pixel of block N, and block L is the upper right pixel of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In frame F602, block E is a block referenced by a motion vector of block J, a block referenced by a motion vector of block K, a block referenced by a motion vector of block L, or a motion of block M Let it be the block that the vector references. At that time, block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, and block C is an upper right pixel of block E. A block including the pixel right above the pixel, a block D including the pixel immediately left of the pixel at the bottom left of the block E, a block F including a pixel right above the pixel at the top right of the block E, a block G Is the block containing the pixel directly below the leftmost pixel of block E, block H is the block containing the pixel directly below the rightmost pixel of block E, block I is the block right below the rightmost pixel of block E A block including the lower pixel.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block J or the block K refers to It is possible that the block referenced by block or block L or the block referenced by block M may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (J, K, L, or M).

  At this time, details of the probability switching signal generator 213 will be described with reference to the flowchart of FIG. The probability switching signal indicates the likelihood of the index indicating the motion vector prediction value determination method, and the entropy coding is performed by switching the probability table based on this signal, so that the appropriate probability corresponding to the data generation probability can be obtained. Indexes can be encoded with a code amount.

  In the probability switching signal generator 213, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 213 moves the encoded blocks A, B, C, D, E, F, G, H, I, J, K, L, and M inputted via the line L212. The reference frame number corresponding to the vector and the reference frame number of the target block input via the line L216 are acquired (step S1101).

  Note that the motion vector acquired in step S1101 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S1101, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Next, there are 4 types of differences: motion vector difference between block A and block E, motion vector difference between block C and block E, motion vector difference between block G and block E, and motion vector difference between block I and block E. The average value of the differences is calculated. In addition, the difference between the motion vector of block B and block E, the difference of the motion vector of block D and block E, the difference of the motion vector of block F and block E, and the difference of the motion vector of block H and block E are 4 ways in total. Calculate the average value of the differences. In addition, the difference between the motion vector of block J and block K, the difference of the motion vector of block J and block M, the difference of the motion vector of block E and block K, and the difference of the motion vector of block E and block M are 4 ways in total. An average value of the differences is calculated (step S1102).

  Subsequently, a combination having the smallest value among the three difference average values calculated in step S1102 is determined. (Step S1103).

  As a result of the determination in step S1103, if the combination that gives the smallest difference average value is block A and E, block C and E, block G and E, block I and E, the motion vector prediction value of block N is block A signal Signal_i = (i = x, y) 0 indicating that it is likely to be determined as an average value of motion vectors of A, B, C, D, E, F, G, H, and I, and The average difference value MViD_mean (i = x, y) at the time is the difference between the motion vectors of block A and block E, the difference between the motion vectors of block C and block E, the difference of the motion vectors of block G and block E, and the block I and It is set as the average value of a total of four types of differences of the motion vector difference of the block E. When the combination that gives the smallest difference average value is block B and E, block D and E, block F and E, block H and E, the motion vector prediction value of block N is block B, D, E, F, A signal Signal_i = 1 (i = x, y) indicating that it is likely to be determined as an average value of the motion vector of H is generated, and the average difference value MViD_mean (i = x, y) at that time is generated in the block B Difference of motion vectors between block E and block E, difference between motion vectors between block D and block E, difference between motion vectors between block F and block E, and difference between motion vectors between block H and block E. And When the combination that gives the smallest difference average value is block J and K, block J and M, block E and K, block E and M, the motion vector prediction value of block N is block J, K, L, M, A signal Signal_i = 2 (i = x, y) indicating that it is likely to be determined as an average value of the motion vector of E is generated, and the average difference value MViD_mean (i = x, y) at that time is generated in the block J Difference between motion vectors of block K and block K, difference of motion vectors of block J and block M, difference of motion vectors of block E and block K, and difference of motion vectors of block E and block M (Step S1104).

  Finally, the index Index_i (i = x, y) and the motion vector average difference value MViD_mean (i = x, y) determined in step S1104 are output (step S1105), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y).

  The series of processes shown in the flowchart of FIG. 21 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  In step S1102, the average difference value is used as information representing the similarity of the motion vectors of each block group. Instead, the variance value or standard deviation of each block group may be used. In that case, the average difference value generated in step S1104 becomes a variance value or a standard deviation.

  In the embodiment shown in the flowchart of FIG. 21, one of the three probability switching signals is selected, but any one of the three probability switching signals may be selected.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 13 may be determined as follows. However, the determination method needs to correspond to the motion predictor 212.

  The target block is N, and the frame F601 is a frame including the block N. Frame F602 is an already encoded frame.

  In frame F601, block J is a block including the pixel immediately above the upper left pixel of block N, block K is a block including the pixel immediately above the upper left pixel of block N, and block L is the upper right pixel of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In frame F602, block E is a block referenced by a motion vector of block J, a block referenced by a motion vector of block K, a block referenced by a motion vector of block L, or a motion of block M Let it be the block that the vector references. At that time, block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, and block C is an upper right pixel of block E. A block including the pixel right above the pixel, a block D including the pixel immediately left of the pixel at the bottom left of the block E, a block F including a pixel right above the pixel at the top right of the block E, a block G Is the block containing the pixel directly below the leftmost pixel of block E, block H is the block containing the pixel directly below the rightmost pixel of block E, block I is the block right below the rightmost pixel of block E A block including the lower pixel.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block J or the block K refers to It is possible that the block referenced by block or block L or the block referenced by block M may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (J, K, L, or M).

  At this time, the entropy encoder 214 switches the occurrence probability distribution table of each index indicating the motion vector prediction value determination method possessed in advance based on a probability switching signal input via the line L215. . Specifically, the occurrence probability is predicted by the index with the highest occurrence probability by Signal_i (i = x, y) and MViD_mean (i = x, y), and the probability table is switched accordingly.

  Finally, a moving picture coding program that causes a computer to operate as the moving picture coding apparatus 20 will be described. FIG. 52 shows a configuration of a moving image encoding program P20 according to the present embodiment, and the moving image encoding program P20 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. Note that the configuration of the computer for executing the program stored in the recording medium is the same as the configuration of FIGS. 63 and 64 already described, and therefore, a duplicate description is omitted.

  As shown in FIG. 52, the moving image encoding program P20 includes a main module P201 that supervises processing, a block division module P202, a transform module P203, a quantization module P204, an inverse quantization module P205, and an inverse transform. Module P206, storage module P207, prediction signal generation module P208, motion prediction module P209, motion detection module P210, addition module P211, subtraction module P212, probability switching signal generation module P213, and entropy encoding module P214.

  Block division module P202, transform module P203, quantization module P204, inverse quantization module P205, inverse transform module P206, prediction signal generation module P208, motion prediction module P209, motion detection module P210, addition module P211, subtraction module P212, probability The switching signal generation module P213 and the entropy encoding module P214 make the computer realize the block divider 201, the converter 207, the quantizer 208, the inverse quantizer 209, the inverse transformer 210, and the prediction signal shown in FIG. The functions of the generator 204, motion predictor 212, motion detector 203, adder 211, subtractor 206, probability switching signal generator 213, and entropy encoder 214 are the same. The functions that the storage module P207 realizes in the computer are the same as the functions of the frame memory 202 and the motion vector memory 205.

[Third Embodiment]
The moving picture encoding apparatus 30 according to the present invention will be described with reference to FIG.

  As shown in FIG. 22, the moving image encoding apparatus 30 includes a block divider 301, a frame memory 302, a motion detector 303, a prediction signal generator 304, and a motion vector memory 305 as functional components. , Subtractor 306, converter 307, quantizer 308, inverse quantizer 309, inverse transformer 310, adder 311, motion predictor 312, subtractor 313, and probability switching signal A generator 314 and an entropy encoder 315 are provided.

  An input video signal (moving image signal) input to the moving image encoding device 30 is composed of a time series of image signals in units of frames (hereinafter referred to as frame image signals).

  The block divider 301 divides the frame image signal to be encoded of the input video signal input via the line L301 into a plurality of regions (blocks), and blocks the block to be encoded via the line L302 ( Hereinafter, the image signal of the target block) is output. In this embodiment, the block is divided into 8 × 8 pixels, but may be divided into other sizes or shapes. The following encoding process is performed for each block.

  The frame memory 302 stores a previously encoded frame image signal (hereinafter referred to as a reference frame image signal) input via the line L303, and outputs it via the line L304 as necessary.

  The motion detector 303 searches an image signal pattern similar to the image signal pattern of the target block input via the line L302 within a predetermined range from the reference frame image signal input via the line L304. Then, a motion vector, which is a spatial displacement amount between both image signal patterns, is detected and output via the line L305. Also, the frame number (hereinafter referred to as reference frame number) referred to by the motion vector is detected and output via the line L316.

  The prediction signal generator 304 includes the motion vector signal value of the target block input via the line L305, the reference frame number input via the line L316, and the reference frame image input via the line L304. A predicted image signal for the target block is generated from the signal and output via a line L306.

  The motion vector memory 305 stores a motion vector input via the line L305 and a reference frame number input via the line L316 together with a motion vector and a reference frame number encoded in the past. Output via line L312 if necessary.

  The subtractor 306 subtracts the prediction image signal for the target block input via the line L306 from the image signal of the target block input via the line L302, generates a residual signal, and generates the line L307. Output via.

  The converter 307 orthogonally transforms the residual signal input via the line L307 and outputs a transform coefficient via the line L308.

  The quantizer 308 quantizes the transform coefficient input via the line L308, and outputs the quantized transform coefficient via the line L309.

  The inverse quantizer 309 inversely quantizes the quantized transform coefficient input via the line L309 and outputs the restored transform coefficient via the line L310.

  The inverse transformer 310 performs inverse orthogonal transform on the transform coefficient input via the line L310, and outputs a residual signal restored via the line L311.

  The adder 311 adds the residual signal input via the line L311 and the prediction image signal for the target block input via the line L306, restores the image signal of the target block, and sets the line L303. Output via.

  The motion predictor 312 receives the motion vector of the previously encoded block input via the line L312 and the reference frame number corresponding thereto, and the reference frame number of the target block input via the line L316. Then, the motion vector prediction value of the target block is calculated and output via the line L313. The detailed operation of the motion predictor 312 is assumed to be the same as that of the motion predictor 112, and the description is omitted.

  The subtractor 313 subtracts the motion vector prediction value of the target block input via the line L313 from the motion vector signal value of the target block input via the line L305 to generate a motion vector difference value. . Then, the motion vector difference value is output via the line L314.

  The probability switching signal generator 314 inputs the motion vector of the previously encoded block input via the line L312 and the corresponding reference frame number, and the reference frame of the target block input via the line L316. From the number, a probability switching signal indicating a probability table to be switched according to a flow to be described later is generated and output via a line L315.

  The entropy encoder 315 encodes the quantized transform coefficient input via the line L309. Also, based on the probability switching signal input via the line L315, the probability table used when entropy encoding is switched by the method described later, and then the motion vector difference value input via the line L314 Is encoded. After the encoded transform coefficient and motion vector difference value are multiplexed, the compressed stream is output to the line L316.

  Switching the probability table used when entropy encoding is performed as described above has an effect of increasing the encoding efficiency by encoding the motion vector difference value with an appropriate code amount based on the appearance frequency.

  The detailed operation of the probability switching signal generator 314 preferably corresponds to the operation of the motion predictor 312. In the following description, the motion vector prediction value is determined from the motion vector of the encoded block, and an index for indicating the motion vector prediction value will not be encoded. However, an index for indicating the motion vector prediction value will be described. The present invention can also be applied to the form of encoding.

  When the operation of the motion predictor 312 performs the operation described with reference to FIGS. 2 and 4 of the motion predictor 112 described above, the detailed operation of the probability switching signal generator 314 is as shown in the schematic diagram of FIG. It is preferable to follow the flowchart of FIG. The description of FIG. 2 is the same as that in the motion predictor 112 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value. By switching the probability table based on this signal and performing entropy coding, the probability switching signal corresponds to the data occurrence probability. The motion vector difference value can be encoded with an appropriate code amount.

  In the probability switching signal generator 314, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 314 passes the reference frame number corresponding to the motion vectors of the encoded blocks A, B, C, D, E, and F inputted via the line L312 and the line L316. The reference frame number of the target block input in step S1201 is acquired.

  Note that the motion vector acquired in step S1201 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S1201, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And four different motion vector differences of block F are calculated (step S1202).

  Next, it is determined which combination of the four differences calculated in step S1202 has the smallest difference. (Step S1203).

  As a result of the determination in step S1203, when the combination that gives the smallest difference is block B and block E, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. A signal Signal_i (i = x, y) = 0 indicating the difference between the motion vectors of the block B and the block E is used as a difference MViD (i = x, y) at that time. When the combination that gives the smallest difference is the block C and the block F, the signal Signal_i = 0 (which indicates that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block A. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block F. When the combination that gives the smallest difference is the block D and the block E, the signal Signal_i = 1 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block E. When the combination that gives the smallest difference is the block D and the block F, the signal Signal_i = 2 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block E. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block F. (Step S1204).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S1204 are output (step S1205), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Alternatively, only MViD (i = x, y) may be used.

  Note that the series of processes shown in the flowchart of FIG. 23 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the entropy encoder 315 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on a probability switching signal input via the line L315. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  When the operation of the motion predictor 312 performs the operation described with reference to FIGS. 6 and 7 of the motion predictor 112 described above, the detailed operation of the probability switching signal generator 314 is as shown in the schematic diagram of FIG. It is preferable to follow the flowchart of FIG. The description of FIG. 6 is the same as that in the motion predictor 112 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value. By switching the probability table based on this signal and performing entropy coding, the probability switching signal corresponds to the data occurrence probability. The motion vector difference value can be encoded with an appropriate code amount.

  In the probability switching signal generator 314, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 314 sets the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, and G input via the line L312 and the line L316. The reference frame number of the target block that is input via is acquired (step S1301).

  Note that the motion vector acquired in step S1301 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the encoded block motion vector does not exist in step S1301, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S1302).

  Next, it is determined which combination of the five differences calculated in step S1302 has the smallest difference. (Step S1303).

  As a result of the determination in step S1303, if the combination that gives the smallest difference is block B and block F, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block B. A signal Signal_i = 0 (i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as a difference value between the motion vectors of the block B and the block F. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D. When the combination that gives the smallest difference is the block E and the block F, the signal Signal_i = 3 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block G. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block E and block F. When the combination that gives the smallest difference is the block E and the block G, the signal Signal_i = 4 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is used as the difference value between the motion vectors of block E and block G. (Step S1304).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S1304 are output (step S705), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Further, only MViD (i = x, y) may be used.

  The series of processes shown in the flowchart of FIG. 24 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and generates a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  Also, the difference calculated in step S1302 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. At that time, the smallest combination is determined in step S1303. In step S1304, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D.

  Further, the difference calculated in step S1302 may be two types, that is, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. At that time, the smallest combination is determined in step S1303. In step S1304, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G.

  The series of processes shown in the flowchart of FIG. 24 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and generates a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the entropy encoder 315 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on a probability switching signal input via the line L315. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  When the operation of the motion predictor 312 is the same as the operation described above with reference to FIGS. 8 and 9 of the motion predictor 112, the detailed operation of the probability switching signal generator 314 is as shown in the schematic diagram of FIG. It is preferable to follow the flowchart of FIG. The description of FIG. 8 is the same as that in the motion predictor 112 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value. By switching the probability table based on this signal and performing entropy coding, the probability switching signal corresponds to the data occurrence probability. The motion vector difference value can be encoded with an appropriate code amount.

  In the probability switching signal generator 314, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 314 includes reference frame numbers corresponding to the motion vectors of the encoded blocks A, B, C, D, E, F, G, H, and I that are input via the line L312. The reference frame number of the target block input via the line L316 is acquired (step S1401).

  Note that the motion vector acquired in step S201 is preferably scaled based on the temporal distance between the frame including the target block and the frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the motion vector of the encoded block does not exist in step S1401, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, variance of motion vector of block D, motion vector of block E and motion vector of block F, motion vector of block G and block Three sets of variances of H motion vector and block I motion vector variance are calculated. (Step S1402)
Next, it is determined which combination of the three variances calculated in step S1402 has the smallest variance. (Step S1403).

  As a result of the determination in step S1403, when the combination that gives the smallest difference is block A, block B, and block C, the motion vector prediction value of block J is the motion vector of block A, the motion vector of block B, and the block A signal Signal_i = 0 (i = x, y) indicating that it is likely that the average value of the motion vector of C is determined is generated, and the variance MVi_var (i = x, y) at that time is generated as a block A It is assumed that the motion vectors of block B and block C are distributed. When the combination that gives the smallest difference is block D, block E, and block F, the motion vector prediction value of block J is the average of the motion vector of block D, the motion vector of block E, and the motion vector of block F. A signal Signal_i = 1 (i = x, y) indicating that it is likely to be determined is generated, and a distribution MVi_var (i = x, y) at that time is used as a motion vector of the block D, the block E, and the block F. Of variance. When the combination that gives the smallest difference is block G, block H, and block I, the motion vector prediction value of block J is the average value of the motion vector of block G, the motion vector of block H, and the motion vector of block I. A signal Signal_i = 2 (i = x, y) indicating that it is likely to be determined is generated, and a distribution MVi_var (i = x, y) at that time is used as a motion vector of the block G, the block H, and the block I Of variance. (Step S1404).

  Finally, the signal values Signal_i (i = x, y) and MVi_var (i = x, y) determined in step S1404 are output (step S1405), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Further, only MVi_var (i = x, y) may be used.

  The series of processes shown in the flowchart of FIG. 25 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  In addition, although variance is used as a value representing the similarity of motion vectors in step S1402, a motion vector difference value between blocks or an average of the plurality of difference values may be used.

  In step S1404, a signal indicating that it is likely that the average value of the three motion vectors is used as the predicted value of the motion vector is generated. The intermediate value of any one of the three motion vectors or any one of the values is used as the predicted value. It may indicate that it is plausible to do.

  In FIG. 25, the probability switching signal is determined by referring to the motion vectors of three blocks that are continuous in the time direction and the spatial direction, but this may be only in the spatial direction. Further, only two blocks or four or more blocks may be used in each direction.

  At this time, the entropy encoder 315 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on a probability switching signal input via the line L315. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD_var (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  Finally, a moving picture coding program that causes a computer to operate as the moving picture coding apparatus 30 will be described. FIG. 53 shows a configuration of a moving image encoding program P30 according to the present embodiment, and the moving image encoding program P30 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. Note that the configuration of the computer for executing the program stored in the recording medium is the same as the configuration of FIGS. 63 and 64 already described, and therefore, a duplicate description is omitted.

  As shown in FIG. 53, the moving image encoding program P30 includes a main module P301 that controls processing, a block division module P302, a transform module P303, a quantization module P304, an inverse quantization module P305, and an inverse transform. Module P306, storage module P307, prediction signal generation module P308, motion prediction module P309, motion detection module P310, addition module P311, subtraction module P312, probability switching signal generation module P313, and entropy encoding module P314.

  Block division module P302, transform module P303, quantization module P304, inverse quantization module P305, inverse transform module P306, prediction signal generation module P308, motion prediction module P309, motion detection module P310, addition module P311, probability switching signal generation module P313 and the function that the entropy encoding module P314 realizes in the computer are the block divider 301, the converter 307, the quantizer 308, the inverse quantizer 309, the inverse transformer 310, the prediction signal generator 304, The functions of the motion predictor 312, motion detector 303, adder 311, probability switching signal generator 314, and entropy encoder 315 are the same. The functions that the storage module P307 implements in the computer are the same as the functions of the frame memory 302 and the motion vector memory 305. The function that the subtraction module P312 realizes in the computer is the same as the function of the subtractors 306 and 313.

[Fourth Embodiment]
A moving picture decoding apparatus 40 according to the present invention will be described with reference to FIG.

  As shown in FIG. 19, the moving picture decoding apparatus 40 includes a data analyzer 401, an inverse quantizer 402, an inverse transformer 403, a frame memory 404, and a motion vector memory 405 as functional components. A motion predictor 406, an adder 407, a prediction signal generator 408, and an adder 409.

  The moving picture decoding apparatus 40 receives the compressed stream encoded by the above-described moving picture encoding apparatus. This compressed stream includes a residual signal that represents the difference between the original signal of the block to be decoded and the prediction signal, and information related to the generation of the prediction signal of the block to be decoded.

  The data analyzer 401 analyzes the compressed stream input via the line L401, thereby extracting a residual signal of a block (hereinafter referred to as a target block) to be subjected to orthogonal transform and quantization. Output via line L402. Also, a motion vector difference value, which is the difference between the motion vector of the target block and its predicted value, is extracted and output via line L403, and the reference frame number of the target block is extracted and output via line L404. To do.

  The inverse quantizer 402 inverse quantizes the quantized transform coefficient input via the line L402 and outputs the restored transform coefficient via the line L405.

  The inverse transformer 403 performs inverse orthogonal transform on the transform coefficient input via the line L405, and outputs the restored residual signal via the line L406.

  The frame memory 404 stores a previously decoded frame image signal (hereinafter referred to as a reference frame image signal) input via the line L407 and outputs it via the line L408 as necessary.

  The motion vector memory 405 includes the motion vector of the target block input via the line L409 and the target input via the line L404 together with the previously decoded motion vector and the corresponding reference frame number. The reference frame number of the block is stored and output via the line L410 as necessary.

  The motion predictor 406 calculates the motion vector prediction value of the target block from the motion vector of the previously decoded block input via the line L410 according to the flow to be described later, and outputs it via the line L411. .

  The adder 407 adds the motion vector prediction value input via the line L411 and the motion vector difference value input via the line L403, restores the motion vector of the target block, and passes the line L409. And output.

  The prediction signal generator 408 receives the motion vector signal value of the target block input via the line L409, the reference frame number of the target block input via the line L404, and the line L408. From the reference frame image signal, a predicted image signal for the target block is generated and output via the line L412.

  The adder 409 adds the residual signal input via the line L406 and the prediction image signal for the target block input via the line L412, restores the image signal of the target block, and sets the line L407 Output via.

  Next, the motion predictor 406 will be described in detail with reference to FIG. FIG. 27 is a schematic diagram for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  In FIG. 27, the target block is H, and the frame F701 is a frame including the block H. A frame F702 is a frame that has already been decoded, and is adjacent to the frame F701 in the time direction.

  In frame F701, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the uppermost leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F702, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block H. Block B is a block including a pixel immediately above the upper leftmost pixel of block A, and block C is a block including a pixel immediately left of the uppermost left pixel of block A.

  FIG. 28 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  First, the motion predictor 406 acquires reference frame numbers corresponding to the motion vectors of decoded blocks A, B, C, E, and F input via the line L410 (step S1501).

  Note that the motion vector acquired in step S1501 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, E, and F motion vectors can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S1501, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, and the difference of the motion vector of block C and the motion vector of block F are calculated (step S1502).

  Next, it is determined whether at least one of the two differences calculated in step S1502 is equal to or less than a threshold value (step S1503).

  If it is determined in step S1503 that at least one of the two differences is equal to or less than the threshold (step S1503: YES), the process proceeds to step S1504. If the two differences are both greater than the threshold (step S1503: NO), the process proceeds to step S1505.

  In step S1504, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  In step S1505, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F.

  Finally, the determined value PMVi (i = x, y) is output (step S1506), and the process ends.

  Note that the determination in step S1503 may determine whether both of the two differences are equal to or less than a threshold value.

  In step S1505, the motion vector prediction value of the block H is set to the same value as the motion vector of the block F, but the motion vector may be a zero vector (that is, the motion vector is not predicted). An arbitrary motion vector determined in advance may be used as the predicted value.

  Further, the motion vector prediction value in the motion predictor 406 may be determined according to the flowchart of FIG.

  In FIG. 29, it is assumed that predicted values are independently determined for the x component and the y component of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 acquires reference frame numbers corresponding to the motion vectors of the decoded blocks A, B, C, D, E, and F input via the line L410 (step S1601).

  Note that the motion vector acquired in step S1601 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S1601, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And 4 types of motion vector differences of block F are calculated (step S1602).

  Next, it is determined which combination of the four differences calculated in step S1602 has the smallest difference. (Step S1603).

  As a result of the determination in step S1603, if the combination that gives the smallest difference is block B and block E, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. In the case of block C and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. In the case of block D and block E, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F. In the case of block D and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block E (step S1604).

  Finally, the value PMVi (i = x, y) determined in step S1604 is output (step S1605), and the process ends.

  The series of processes shown in the flowchart of FIG. 29 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and determines the x component and the y component of the motion vector at the same time. May be.

  The procedure for determining the motion vector prediction value in the motion predictor 406 may follow the flowchart described in FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 acquires reference frame numbers corresponding to motion vectors of decoded blocks A, B, C, D, E, and F input via the line L410 (step S1701).

  Note that the motion vector acquired in step S1701 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S1701, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block D and the motion vector of block E, and the difference between the motion vector of block D and the motion vector of block F are calculated. If the former is larger (step S1702: YES), the process is performed. The process proceeds to S1703. If this is not the case (step S1702: NO), the process proceeds to step S1704 (step S1702).

  In step S1703, the difference between the motion vector of block E and the motion vector of block B, and the difference of the motion vector of block D and the motion vector of block F are calculated. If the former is larger (step S1703: YES), the process is performed. The process proceeds to step S1705. If not (NO at step S1703), the process proceeds to step S1706.

  In step S1704, the difference between the motion vector of block C and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block A are calculated. If the former is larger (step S1704: YES), the process is performed. The process proceeds to step S1707. If not (NO at step S1704), the process proceeds to step S1708.

  In step S1705, the motion vector prediction value of block H is determined to be the same value as the motion vector of block E.

  In step S1706, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  In step S1707, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F.

  In step S1708, the motion vector prediction value of block H is determined to be the same value as the motion vector of block A.

  Finally, the determined motion vector prediction value PMVi (i = x, y) is output (step S1709), and the process ends.

  The series of processes shown in the flowchart of FIG. 30 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 27 may be determined as follows. The procedure for determining the motion vector prediction value at that time may follow any of the flowcharts described in FIG. 28, FIG. 29, or FIG.

  The target block is H, and the frame F701 is a frame including the block H. A frame F702 is a frame that has already been decoded.

  In frame F701, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the uppermost leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F702, the block A is a block that is referenced by the motion vector of the block D, a block that is referenced by the motion vector of the block E, or a block that is referenced by the motion vector of the block F. Also in this case, the block B is a block including a pixel immediately above the uppermost leftmost pixel of the block A, and the block C is a block including a pixel rightward of the uppermost leftmost pixel of the block A.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block D or the block E refers to It is conceivable that the block or the block referenced by block F may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these blocks or the block with the highest ratio in the spanning area The motion vector is used as a motion vector of a block referred to by the corresponding block (D, E, or F).

  Further, the motion vector prediction value in the motion predictor 406 may be determined as follows.

  FIG. 31 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  In FIG. 31, the target block is H, and the frame F801 is a frame including the block H. The frame F802 is a frame that has already been decoded, and the frame F801 is a frame that is adjacent in the time direction. The frame F803 is a frame that has already been decoded, and the frame F802 is a frame that is adjacent in the time direction.

  In frame F801, block E is a block including the pixel immediately above the upper left pixel of block H, block F is a block including the pixel immediately above the upper left pixel of block H, and block G is the upper left pixel of block H. A block including a pixel immediately to the left of the pixel.

  In frame F802, block D is a block including a pixel at the same spatial position as the upper leftmost pixel of block H, block B is a block including a pixel immediately above the uppermost left pixel of block D, and block C is the highest block of block D. It is assumed that the block includes a pixel immediately to the left of the upper left pixel.

  In the frame F803, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block D.

  FIG. 32 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  In FIG. 32, it is assumed that predicted values are independently determined for the x component and the y component of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 acquires reference frame numbers corresponding to the motion vectors of the decoded blocks A, B, C, D, E, F, and G input via the line L410 (step S1801). .

  Note that the motion vector acquired in step S1801 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S1801, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S1802).

  Next, it is determined which combination of the five differences calculated in step S1802 has the smallest difference. (Step S1803).

  As a result of the determination in step S1803, when the combination that gives the smallest difference is the block B and the block F, the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block B. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C. In the case of block A and block D, the motion vector prediction value of block H is determined to be the same value as the motion vector of block D. In the case of block E and block F, the motion vector prediction value of block H is determined to be the same value as the motion vector of block G. In the case of block E and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block F. (Step S1804).

  Finally, the value PMVi (i = x, y) determined in step S1804 is output (step S1805), and the process ends.

  The series of processes shown in the flowchart of FIG. 32 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Also, the difference calculated in step S1802 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. At that time, the smallest combination is determined in step S1803. In step S1804, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. Determined by value. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C. In the case of block A and block D, the motion vector prediction value of block H is determined to be the same value as the motion vector of block D.

  Further, the difference calculated in step S1802 may be two types, that is, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. At that time, the smallest combination is determined in step S1803. In step S1804, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. Determined by value. In the case of block C and block G, the motion vector prediction value of block H is determined to be the same value as the motion vector of block C.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 31 may be determined as follows.

  The target block is H, and the frame F801 is a frame including the block H. Frame F802 is a frame that has already been decoded. Frame F803 is a frame that has already been decoded.

  In frame F801, block E is a block including the pixel immediately above the upper left pixel of block H, block F is a block including the pixel immediately above the upper left pixel of block H, and block G is the upper left pixel of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F802, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In frame F803, block A is a block that is referenced by the motion vector of block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  In addition, the motion vector predictor 406 determines the motion vector prediction value of a block that is spatially separated by one or more pixels as shown below, or a block in a frame that is separated by one or more frames in the time direction. It may be performed based on a motion vector.

  Such a method for determining a motion vector prediction value has an effect of obtaining a motion vector prediction value with high accuracy by performing prediction reflecting a large number of motion vectors. It also has the effect of countering the effects of noise and local changes.

  In addition, the motion vector prediction value in the motion predictor 406 is determined as a motion vector prediction value as described below, instead of using the motion vector of a previously decoded block as it is as a motion vector prediction value. One of them may be selected and determined according to the determination method.

  FIG. 33 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  In FIG. 33, the target block is J, and the frame F901 is a frame including the block J. The frame F902 is a frame that has already been decoded, and the frame F901 is a frame that is adjacent in the time direction. The frame F903 is a frame that has already been decoded, and the frame F902 is a frame that is adjacent in the time direction. The frame F904 is a frame that has already been decoded, and the frame F903 is a frame that is adjacent in the time direction.

  In frame F901, block I is a block including the pixel immediately above the uppermost leftmost pixel of block J, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block J, a block F including the pixel immediately left of the uppermost leftmost pixel of the block J, a block E including a pixel immediately above the leftmost pixel of the block F, and a block Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E.

  In the frame F902, the block C is a block including a pixel at the same spatial position as the upper left pixel of the block J.

  In the frame F903, the block B is a block including a pixel at the same spatial position as the upper left pixel of the block C.

  In the frame F904, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block B.

  FIG. 34 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 406. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 receives the reference frame number corresponding to the motion vector of the decoded blocks A, B, C, D, E, F, G, H, and I input via the line L410 and the line L115. The reference frame number of the target block input via is acquired (step S1901).

  Note that the motion vector acquired in step S1901 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S1901, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, variance of motion vector of block D, motion vector of block E and motion vector of block F, motion vector of block G and block Three sets of variances of H motion vector and block I motion vector variance are calculated. (Step S1902)
Next, it is determined which combination of the three variances calculated in step S1902 has the smallest variance. (Step S1903).

  As a result of the determination in step S1903, when the combination that gives the smallest difference is block A, block B, and block C, the motion vector prediction value of block J is the motion vector of block A, the motion vector of block B, and the block The average value of C motion vectors is determined. In the case of block D, block E, and block F, the motion vector prediction value of block J is determined as the average value of the motion vector of block D, the motion vector of block E, and the motion vector of block F. In the case of block G, block H, and block I, the motion vector prediction value of block J is determined as the average value of the motion vector of block G, the motion vector of block H, and the motion vector of block I (step S1904). .

  Finally, the value PMVi (i = x, y) determined in step S1904 is output (step S1905), and the process ends.

  The series of processes shown in the flowchart of FIG. 34 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Further, although variance is used as a value representing the similarity of motion vectors in step S1902, standard deviation, a difference value of motion vectors between blocks, or an average of a plurality of difference values may be used.

  In step S1904, the average value of the three motion vectors is used as the predicted value of the motion vector. However, an intermediate value of the three motion vectors or any one of the values may be used as the predicted value.

  In FIG. 34, the motion vector prediction value is determined with reference to three blocks of motion vectors that are continuous in the time direction and the spatial direction, but this may be only in the time direction. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  Further, in order to increase the prediction accuracy when the region including the target block is moving greatly, the positional relationship of the blocks shown in FIG. 33 may be determined as follows.

  The target block is J, frame F901 is a frame including block J, and frames F902, F903, and F904 are already decoded frames.

  In frame F901, block I is a block including the pixel immediately above the uppermost leftmost pixel of block J, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block J, a block F including the pixel immediately left of the uppermost leftmost pixel of the block J, a block E including a pixel immediately above the leftmost pixel of the block F, and a block Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E.

  In the frame F902, a block C is a block including a pixel referred to by a motion vector of the block I, a block referred to by a motion vector of the block F, or a block including a pixel at the upper left of the upper left pixel of the block J (block K)) is referred to by the motion vector.

  In the frame F903, the block B is a block referred to by the motion vector of the block C.

  In frame F904, block A is a block that is referenced by the motion vector of block B.

  If the above method is used, block I, block F, block K, block C, block block when detecting the position of a similar image signal pattern with accuracy below the size of each block and detecting the motion vector It is possible that the block referenced by B spans multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the motion vector of the block with the highest ratio in the spanning area is applicable. This block is used as a motion vector of a block referred to by a block (I or F or K or C or B).

  Furthermore, the motion vector predictor 406 determines the motion vector prediction value of a block in a block that is spatially separated by one or more pixels, as shown below, or in a frame that is separated by one or more frames in the temporal direction. It may be performed based on a motion vector.

  Such a method for determining a motion vector prediction value has an effect of obtaining a motion vector prediction value with high accuracy by performing prediction reflecting a large number of motion vectors. It also has the effect of countering the effects of noise and local changes.

  In addition, the motion vector prediction value in the motion predictor 406 is determined as a motion vector prediction value as described below, instead of using the motion vector of a previously decoded block as it is as a motion vector prediction value. One of them may be selected and determined according to the determination method.

  FIG. 35 is a schematic diagram used for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  In FIG. 35, the target block is M, and a frame F1001 is a frame including the block M. A frame F1002 is a frame that has already been decoded, and is adjacent to the frame F1001 in the time direction. The frame F1003 is a frame that has already been decoded, and the frame F1002 is a frame that is adjacent in the time direction. The frame F1004 is a frame that has already been decoded, and the frame F1003 is a frame that is adjacent in the time direction.

  In frame F1001, block I is a block including the pixel immediately above the uppermost leftmost pixel of block M, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block M, a block F including the pixel immediately left of the leftmost pixel of the block M, and a block E including a pixel immediately left of the uppermost left pixel of the block F Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E. Also, block K is a block including the pixel immediately above the upper left pixel of block M, block L is a block including the pixel immediately above the upper left pixel of block K, and block J is the upper right pixel of block M. A block including a pixel on the right upper side of.

  In the frame F1002, the block C is a block including a pixel at the same spatial position as the upper left pixel of the block M.

  In the frame F1003, the block B is a block including a pixel at the same spatial position as the upper left pixel of the block C.

  In the frame F1004, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block B.

  FIG. 36 is a flowchart for explaining a procedure for determining a motion vector prediction value in the motion predictor 406.

  First, the motion predictor 406 corresponds to the motion vectors of the decoded blocks A, B, C, D, E, F, G, H, I, J, K, and L that are input via the line L410. A reference frame number is acquired (step S2001).

  Note that the motion vector acquired in step S2001 is preferably scaled based on the temporal distance between the frame including the target block and the frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If there is no decoded block itself or a motion vector of the decoded block in step S2001, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, the variance of the x component of the motion vector of block A, the motion vector of block B, and the motion vector of block C, the variance of the x component of the motion vector of block D, the motion vector of block E, and the motion vector of block F, block The three sets of x component variances of the J motion vector, the block K motion vector, and the block I motion vector variance are calculated. (Step S2002)
Next, it is determined which combination of the three variances calculated in step S2002 has the smallest variance. (Step S2003).

  As a result of the determination in step S2003, if the combination that gives the smallest difference is block A, block B, and block C, the x component of the motion vector prediction value of block M is the motion vector of block A and the motion of block B. The average value of the x component of the vector and the motion vector of block C is determined. In the case of block D, block E, and block F, the x component of the motion vector prediction value of block M is determined as the average value of the motion vector of block D, the motion vector of block E, and the x component of the motion vector of block F Is done. In the case of block J, block K, and block I, the x component of the motion vector prediction value of block M is determined as the average value of the motion vector of block J, the motion vector of block K, and the x component of the motion vector of block I (Step S2004).

Subsequently, variance of y component of motion vector of block A, motion vector of block B and motion vector of block C, variance of y component of motion vector of block G, motion vector of block H and motion vector of block I, block Three sets of variances of the y component of the L motion vector, the motion vector of the block K, and the variance of the y component of the motion vector of the block F are calculated. (Step S2005)
Next, it is determined which combination of the three variances calculated in step S2005 has the smallest variance. (Step S2006).

  As a result of the determination in step S2006, when the combination that gives the smallest difference is block A, block B, and block C, the y component of the motion vector prediction value of block M is the motion vector of block A and the motion of block B. The average value of the y component of the vector and the motion vector of block C is determined. In the case of block G, block H, and block I, the y component of the motion vector prediction value of block M is determined as the average value of the y component of the motion vector of block G, the motion vector of block H, and the motion vector of block I Is done. In the case of block L, block K, and block F, the y component of the motion vector prediction value of block M is determined as the average value of the motion vector of block L, the motion vector of block K, and the y component of the motion vector of block F (Step S2007).

  Finally, the value PMVi (i = x, y) determined in step S2007 is output (step S2008), and the process ends.

  Further, in step S2002 and step S2005, variance is used as a value representing the similarity of motion vectors. However, a standard deviation, a difference value of motion vectors between blocks, or an average of a plurality of difference values may be used. .

  In step S2004 and step S2007, the average value of the three motion vectors is used as the predicted value of the motion vector, but an intermediate value of the three motion vectors or any one of the values may be used as the predicted value.

  In FIG. 36, the motion vector prediction value is determined by referring to three blocks of motion vectors that are continuous in the time direction and the space direction, but only the space direction may be used. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  Further, in order to increase the prediction accuracy when the region including the target block is greatly moved, the positional relationship of the blocks shown in FIG. 35 may be determined as follows.

  The target block is M, the frame F1001 is a frame including the block M, and the frames F1002, F1003, and F1004 are already decoded frames.

  In frame F1001, block I is a block including the pixel immediately above the uppermost leftmost pixel of block M, block H is a block including the pixel immediately above the uppermost leftmost pixel of block I, and block G is the uppermost leftmost of block H. A block including the pixel immediately above the leftmost pixel of the block M, a block F including the pixel immediately left of the leftmost pixel of the block M, and a block E including a pixel immediately left of the uppermost left pixel of the block F Let D be a block including the pixel immediately to the left of the upper leftmost pixel of block E. Also, block K is a block including the pixel immediately above the upper left pixel of block M, block L is a block including the pixel immediately above the upper left pixel of block K, and block J is the upper right pixel of block M. A block including a pixel on the right upper side of.

  In the frame F1002, the block C is a block that is referenced by the motion vector of the block I, a block that is referenced by the motion vector of the block K, or a block that is referenced by the motion vector of the block F.

  In the frame F1003, the block B is a block referred to by the motion vector of the block C.

  In the frame F1004, the block A is a block referred to by the motion vector of the block B.

  If the above method is used, block I, block K, block F, block C, block block when detecting the position of a similar image signal pattern with accuracy below the size of each block and detecting the motion vector It is possible that the block referenced by B spans multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the motion vector of the block with the highest ratio in the spanning area is applicable. This block is used as a motion vector of a block to be referenced by a block (I or K or F or C or B).

  By the way, when calculating a motion vector prediction value, a peripheral motion vector is effectively used by weighting as follows, instead of selecting a part of motion vectors from the motion vectors decoded in the past in the vicinity. A method may be adopted. In this case, by effectively using surrounding motion vectors by weighting, it is possible to reduce the influence of local motion vector changes, obtain a predicted value reflecting the majority of motion vector changes, and improve prediction accuracy. be able to.

  In the following, the peripheral motion vector using method by weighting is applied to (a) an example applied to the block of FIG. 27, (b) an example applied to the block of FIG. 31, and (c) an example applied to the block of FIG. These will be described in order.

  First, (a) as an example applied to the block of FIG. 27, for example, the flowchart of FIG. That is, the procedure for determining the motion vector prediction value in the motion predictor 406 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 acquires reference frame numbers corresponding to motion vectors of decoded blocks A, B, C, D, E, and F input via the line L410 (step S3301).

  Note that the motion vector acquired in step S3301 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S3301, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block E, the difference MViD — 1 (i = x, y) between the motion vector of block C and the motion vector of block F, A difference MViD — 2 (i = x, y) between the motion vector and the motion vector of block E, and a difference MViD — 3 (i = x, y) between the motion vector of block D and the motion vector of block F are calculated (step S3302).

  Next, based on the above equation (2) using the function F having each difference as a variable calculated in step S3302, all the motion vectors of the blocks A, F, and E are effectively used, and a motion vector prediction value is obtained. PMVi (i = x, y) is calculated (step S3303).

  Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3304), and the process ends.

  The series of processes shown in the flowchart of FIG. 60 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be.

  Moreover, as the function F used by Formula (2), you may employ | adopt the function which calculates the reciprocal number of a variable (difference), for example. In this case, for example, when the difference MViD — 0 (i = x, y) between the motion vector of the block B and the motion vector of the block E is small, the reciprocal of the difference, that is, a large value becomes the weighting coefficient of the motion vector of the block A. The weight of the motion vector of block A can be increased.

  Next, (b) as an example applied to the block of FIG. 31, for example, the flowchart of FIG. That is, the procedure for determining the motion vector prediction value in the motion predictor 406 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 406 acquires reference frame numbers corresponding to the motion vectors of the decoded blocks A, B, C, D, E, F, and G input via the line L410 (step S3401). .

  Note that the motion vector acquired in step S3401 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S3401, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block F, the difference MViD — 1 (i = x, y) between the motion vector of block C and the motion vector of block G, The difference MViD — 2 (i = x, y) between the motion vector and the motion vector of block D, the difference MViD — 3 (i = x, y) between the motion vector of block E and the motion vector of block F, the motion vector of block E and the block G Five types of motion vector differences MViD — 4 (i = x, y) are calculated (step S3402).

  Next, all the motion vectors of the blocks B, C, D, G, and F are effectively used based on the above equation (3) using the function F having each difference as a variable calculated in step S3402. A motion vector prediction value PMVi (i = x, y) is calculated (step S3403).

  Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3404), and the process ends.

  The series of processes shown in the flowchart of FIG. 61 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be. Moreover, as the function F used by Formula (3), you may employ | adopt the function which calculates the reciprocal number of a variable (difference), for example. In this case, for example, when the difference MViD — 0 (i = x, y) between the motion vector of the block B and the motion vector of the block F is small, the reciprocal of the difference, that is, a large value becomes the weighting coefficient of the motion vector of the block B. The weight of the motion vector of block B can be increased.

  The difference calculated in step S3402 is the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block F, and the difference MViD — 1 (i = x) between the motion vector of block C and the motion vector of block G. , y). At this time, in step S3403, a function F (MV iD — 2 (i = x, y)), a function F (MV iD — 3 (i = x, y)), and a function F (MV iD — 4 (i = x, y)) are set. The motion vector prediction value PMVi (i = x, y) may be calculated with 0). In this case, the motion vector prediction value PMVi (i = x, y) can be calculated based on only the motion vector difference in the temporal direction without including the motion vector difference in the spatial direction.

  Further, the differences calculated in step S3402 are the difference MViD — 0 (i = x, y) between the motion vector of block B and the motion vector of block F, and the difference MViD — 1 (i = x) of the motion vector of block C and the motion vector of block G. , y), the difference MViD_2 (i = x, y) between the motion vector of the block A and the motion vector of the block D may be used. At this time, in step S3403, the function F (MV iD — 3 (i = x, y)) and the function F (MV iD — 4 (i = x, y)) using the difference not calculated as variables are set to 0 and the motion vector prediction value PMVi (i = x, y) may be calculated. In this case, the motion vector prediction value PMVi (i = x, y) can be calculated based only on the temporal motion vector difference including the motion vector difference between the blocks one frame or more away from the target block. it can.

  Next, (c) as an example applied to the block of FIG. 33, there is a flowchart of FIG. That is, the procedure for determining the motion vector prediction value in the motion predictor 406 may be executed according to the flowchart of FIG. It is assumed that predicted values are determined independently for the motion vector x component and the y component. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion estimator 406 receives the reference frame number corresponding to the motion vector of the decoded blocks A, B, C, D, E, F, G, H, and I input via the line L410 and the line L404. The reference frame number of the target block input via is acquired (step S3501).

  Note that the motion vector acquired in step S3501 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S3501, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, MViV_0 (i = x, y), variance of motion vector of block D, motion vector of block E and motion vector of block F Three sets of variances are calculated: MViV_1 (i = x, y), block G motion vector, block H motion vector, and block I motion vector variance MViV_2 (i = x, y). (Step S3502)
Next, based on the above equation (4) using the function F having each variance as a variable calculated in step S3502, all nine motion vectors of the blocks A to I are effectively used, and a motion vector prediction value is obtained. PMVi (i = x, y) is calculated (step S3503).

  Finally, the calculated motion vector prediction value PMVi (i = x, y) is output (step S3504), and the process ends.

  The series of processes shown in the flowchart of FIG. 62 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and simultaneously determines the x component and the y component of the motion vector. May be. Moreover, as the function F used by Formula (4), you may employ | adopt the function which calculates the reciprocal number of a variable (dispersion), for example. In this case, for example, when the variance MViV — 0 (i = x, y) of the motion vector of block A, the motion vector of block B, and the motion vector of block C is small, the reciprocal of the variance, that is, a large value is the motion vector of block A Since the weighting coefficient of the average value of the motion vector of block B and the motion vector of block C can be increased, the weight of the average value of the motion vector of block A, the motion vector of block B, and the motion vector of block C can be increased. .

  Further, although variance is used as a value representing the similarity of motion vectors in step S3502, a standard deviation, a difference value of motion vectors between blocks, or an average of the plurality of difference values may be used.

  In step S3503, the average value of the three motion vectors is used as the predicted value of the motion vector, but an intermediate value of the three motion vectors or any one of the values may be used as the predicted value.

  In FIG. 62, the motion vector prediction value is determined by referring to three blocks of motion vectors that are continuous in the time direction and the space direction, but this may be only in the time direction. Further, only two blocks or four or more blocks may be used in each direction. Further, a block of frames adjacent in the time direction and a block separated by one or more blocks in the spatial direction may be used in combination.

  The variance calculated in step S3502 is the variance MViV_1 (i = x, y) of the motion vector of block D, the motion vector of block E, and the motion vector of block F, the motion vector of block G, and the motion vector of block H. Only two sets of block vector motion vector variance MViV — 2 (i = x, y) may be used. At this time, in step S3503, the motion vector prediction value PMVi (i = x, y) may be calculated with the function F (MViV — 0 (i = x, y)) having the variance not calculated as 0 as 0. In this case, the motion vector prediction value PMVi (i = x, y) can be calculated only from the spatial direction including a plurality of blocks separated by one block or more from the target block.

  Finally, a video decoding program that causes a computer to operate as the video decoding device 40 will be described. FIG. 54 shows a configuration of a moving picture decoding program P40 according to the present embodiment, and the moving picture decoding program P40 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. Note that the configuration of the computer for executing the program stored in the recording medium is the same as the configuration of FIGS. 63 and 64 already described, and therefore, a duplicate description is omitted.

  As shown in FIG. 54, the moving picture decoding program P40 includes a main module P401 that supervises processing, a data analysis module P402, an inverse quantization module P403, an inverse transform module P404, a storage module P405, and a prediction signal. A generation module P406, a motion prediction module P407, and an addition module P408 are provided.

  The functions that the data analysis module P402, inverse quantization module P403, inverse transform module P404, prediction signal generation module P406, and motion prediction module P407 realize in the computer are the data analyzer 401, inverse quantizer 402, and inverse function of FIG. The functions of the converter 403, the prediction signal generator 408, and the motion predictor 406 are the same. The functions realized by the storage module P405 in the computer are the same as the functions of the frame memory 404 and the motion vector memory 405. The function that the adder module P408 realizes in the computer is the same as the function of the adders 407 and 409.

[Fifth Embodiment]
The moving picture decoding apparatus 50 according to the present invention will be described with reference to FIG.

  As shown in FIG. 37, the moving picture decoding apparatus 50 includes a data analyzer 501, an inverse quantizer 502, an inverse transformer 503, a frame memory 504, and a motion vector memory 505 as functional components. A motion predictor 506, an adder 507, a prediction signal generator 508, an adder 509, and a probability switching signal generator 510.

  The moving picture decoding apparatus 50 receives the compressed stream encoded by the moving picture encoding apparatus described above. This compressed stream includes a residual signal that represents the difference between the original signal of the block to be decoded and the prediction signal, and information related to the generation of the prediction signal of the block to be decoded.

  The data analyzer 501 analyzes the compressed stream input via the line L501 to extract a residual signal of a block (hereinafter referred to as a target block) to be decoded and subjected to direct transform and quantization. Output via line L502. Also, a motion vector difference value, which is the difference between the motion vector of the target block and its predicted value, is extracted and output via line L503, and the reference frame number of the target block is extracted and output via line L504. To do. Also, based on the probability switching signal input via the line L512, the probability table used for data extraction is switched by the method described later, and then an index indicating the motion vector prediction value of the target block is extracted, and the line L514 Output via.

  The inverse quantizer 502 inversely quantizes the quantized transform coefficient input via the line L502 and outputs the restored transform coefficient via the line L505.

  The inverse transformer 503 performs inverse orthogonal transform on the transform coefficient input via the line L505, and outputs a residual signal restored via the line L506.

  The frame memory 504 stores a previously decoded frame image signal (hereinafter referred to as a reference frame image signal) input via the line L507 and outputs it via the line L508 if necessary.

  The motion vector memory 505 includes the motion vector of the target block input via the line L509 and the target input via the line L504, together with the previously decoded motion vector and the corresponding reference frame number. The reference frame number of the block is stored and output via the line L510 as necessary.

  The motion estimator 506 applies an index according to a flow described later from an index indicating a motion vector prediction value input via the line L514 and a motion vector of a previously decoded block input via the line L510. The block motion vector prediction value is calculated and output via the line L511.

  The adder 507 adds the motion vector prediction value input via the line L511 and the motion vector difference value input via the line L503, restores the motion vector of the target block, and passes the line L509. And output.

  The prediction signal generator 508 receives the motion vector signal value of the target block input via the line L509, the reference frame number of the target block input via the line L504, and the line L508. From the reference frame image signal, a predicted image signal for the target block is generated and output via the line L513.

  The adder 509 adds the residual signal input via the line L506 and the predicted image signal for the target block input via the line L513, restores the image signal of the target block, and sets the line L507 Output via.

  The probability switching signal generator 510 receives the motion vector of the previously encoded block input via the line L510, the corresponding reference frame number, and the reference frame of the target block input via the line L504. From the number, a probability switching signal indicating a probability table to be switched according to a flow to be described later is generated and output via a line L512.

  Next, the motion predictor 506 will be described in detail with reference to FIG. FIG. 38 is a schematic diagram for explaining a procedure for restoring a motion vector prediction value from an index indicating a motion vector prediction value in the motion predictor 506.

  In FIG. 38, the target block is H, and the frame F1101 is a frame including the block H. The frame F1102 is a frame that has already been decoded, and is adjacent to the frame F1101 in the time direction.

  In frame F1101, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the upper leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F1102, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block H. Block B is a block including a pixel immediately above the upper leftmost pixel of block A, and block C is a block including a pixel immediately left of the uppermost left pixel of block A.

  FIG. 39 is a flowchart for explaining a procedure for restoring a motion vector prediction value from an index indicating a motion vector prediction value in the motion predictor 506. Assume that predicted values are independently determined for the x and y components of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

First, the motion predictor 506 acquires an index indicating the motion vector prediction value of the target block via the line L514. (Step S2101).
Next, the index acquired in step S2101 is determined (step S2102).

  When the index is 0 as a result of the determination in step S2102, the motion vector of the block A that has been encoded in the past and input via the line L510 is set as a motion vector prediction value, and the index is 1. Uses the motion vector of the block E encoded in the past input via the line L510 as a motion vector prediction value, and if the index is 2, the index is stored in the past input via the line L510. The motion vector of the converted block F is set as a motion vector prediction value (step S2103).

  Note that the motion vector acquired in step S2103 is preferably scaled based on the temporal distance between the frame including the target block and the frame referenced by the motion vector of the target block. Since the time of the frame referenced by the motion vector of each block is used for scaling, the reference frame number corresponding to the previously encoded block input via the line L510 and the line L504 are used. It is necessary to obtain the time from the reference frame number of the input target block.

  The obtained decoded blocks A and E based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , The magnitude of the motion vector of F can be corrected. As a specific example, if the time of the frame referred to by the motion vector MViA (i = x, y) of the block A is ta and the time of the frame including the block A is t1, the scaled motion vector MViA ′ ( i = x, y) is a vector given by equation (1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  Finally, the motion vector prediction value PMVi (i = x, y) determined in step S2103 is output (step S2105), and the process ends.

  Note that the series of processes shown in the flowchart of FIG. 29 may be determined simultaneously using an index common to the x and y components without being divided into the x and y components.

  Further, in order to increase the prediction accuracy when the region including the target block is moving greatly, the positional relationship of the blocks shown in FIG. 38 may be determined as follows.

  The target block is H, and the frame F1101 is a frame including the block H. A frame F1102 is an already encoded frame.

  In frame F1101, block D is a block including the pixel immediately above the upper leftmost pixel of block H, block E is a block including the pixel immediately above the uppermost left pixel of block H, and block F is the upper leftmost of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F1102, the block A is a block that is referenced by the motion vector of the block D, a block that is referenced by the motion vector of the block E, or a block that is referenced by the motion vector of the block F. Also in this case, the block B is a block including a pixel immediately above the uppermost leftmost pixel of the block A, and the block C is a block including a pixel rightward of the uppermost leftmost pixel of the block A.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block D or the block E refers to It is conceivable that the block or the block referenced by block F may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these blocks or the block with the highest ratio in the spanning area The motion vector is used as a motion vector of a block referred to by the corresponding block (D, E, or F).

  At this time, details of the probability switching signal generator 510 will be described with reference to the flowchart of FIG. The probability switching signal indicates the likelihood of the occurrence probability of the index indicating the motion vector prediction value, and the probability is the same as the probability table used when the probability table is switched based on this signal and the data is encoded. Data can be extracted by performing data analysis using a table.

  Therefore, the encoding device and the decoding device have a common probability table in advance, and it is necessary to switch the probability table by a common mechanism.

  In the probability switching signal generator 510, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 510 receives the reference frame numbers corresponding to the motion vectors of the decoded blocks A, B, C, D, E, and F inputted via the line L510, and the line L504. The reference frame number of the target block input in step S2201 is acquired.

  Note that the motion vector acquired in step S2201 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S2201, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And four different motion vector differences of block F are calculated (step S2202).

  Next, it is determined which combination of the four differences calculated in step S2202 has the smallest difference. (Step S2203).

  As a result of the determination in step S2203, when the combination that gives the smallest difference is block B and block E, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. A signal Signal_i (i = x, y) = 0 indicating the difference between the motion vectors of the block B and the block E is used as a difference MViD (i = x, y) at that time. When the combination that gives the smallest difference is the block C and the block F, the signal Signal_i = 0 (which indicates that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block A. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block F. When the combination that gives the smallest difference is the block D and the block E, the signal Signal_i = 1 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block E. When the combination that gives the smallest difference is the block D and the block F, the signal Signal_i = 2 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block E. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block F. (Step S2204).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S2204 are output (step S2205), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y).

  Note that the series of processes shown in the flowchart of FIG. 40 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the data analyzer 501 switches a table of occurrence probability distribution of each index indicating a plurality of motion vector prediction values possessed in advance based on a probability switching signal input via the line L512. Specifically, the index with the highest probability of occurrence is calculated by Signal_i (i = x, y), the probability of occurrence is predicted by MViD (i = x, y), and switching to the probability table is performed accordingly.

  Further, the motion predictor 506 may operate as follows. The motion predictor 506 will be described in detail with reference to FIG. FIG. 41 is a schematic diagram for explaining a procedure for restoring a motion vector prediction value from an index indicating a motion vector prediction value in the motion predictor 506.

  In FIG. 41, the target block is H, and the frame F1201 is a frame including the block H. A frame F1202 is an already encoded frame, and is adjacent to the frame F1201 in the time direction. A frame F1203 is an already encoded frame, and is adjacent to the frame F1202 in the time direction.

  In frame F1201, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost left of block H. A block including a pixel immediately to the left of the pixel.

  In frame F1202, block D is a block including a pixel at the same spatial position as the upper left pixel of block H, block B is a block including a pixel immediately above the upper left pixel of block D, and block C is the highest block of block D. It is assumed that the block includes a pixel immediately to the left of the upper left pixel.

  In the frame F1203, the block A is a block including a pixel at the same spatial position as the upper left pixel of the block D.

  FIG. 42 is a flowchart for explaining a procedure for restoring a motion vector prediction value from an index indicating a motion vector prediction value in the motion predictor 506. Assume that predicted values are independently determined for the x and y components of the motion vector. Specifically, first, the predicted value of the x component of the motion vector is determined and output, and then the predicted value of the y component is determined and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 506 acquires an index indicating the motion vector prediction value of the target block via the line L514. (Step S2301).

  Next, the index acquired in step S2301 is determined (step S2302).

  If the index is 0 as a result of the determination in step S2302, the motion vector of the previously encoded block B input via the line L510 is used as a motion vector prediction value, and the index is 1. Uses the motion vector of the block C encoded in the past input via the line L510 as a motion vector prediction value, and if the index is 2, the code is encoded in the past input via the line L510. If the motion vector of the already-encoded block D is a motion vector prediction value and the index is 3, the motion vector of the previously encoded block G input via the line L510 is the motion vector prediction value. If the index is 4, the motion vector of the previously encoded block E input via the line L510 is moved. Vector prediction value (step S2303).

  Note that the motion vector acquired in step S2303 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referenced by the motion vector of each block is used for scaling, the reference frame number corresponding to the previously encoded block input via the line L510 and the line L504 are used. It is necessary to obtain the time from the reference frame number of the input target block.

  The obtained decoded blocks A and E based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , The magnitude of the motion vector of F can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  Finally, the motion vector prediction value PMVi (i = x, y) determined in step S2303 is output (step S2304), and the process ends.

  Note that the series of processes shown in the flowchart of FIG. 42 may be simultaneously determined using an index common to the x component and the y component without being divided into the x component and the y component.

  In addition, in order to improve the prediction accuracy when the region including the target block has moved greatly, the positional relationship of the blocks shown in FIG. 41 may be determined as follows.

  The target block is H, and the frame F1201 is a frame including the block H. A frame F1202 is an already encoded frame. The frame F1203 is an already encoded frame.

  In frame F1201, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost left of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F1202, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In the frame F1203, the block A is a block that is referenced by the motion vector of the block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  At this time, details of the probability switching signal generator 510 will be described with reference to the flowchart of FIG.

  The probability switching signal indicates the likelihood of the occurrence probability of the index indicating the motion vector prediction value, and the probability is the same as the probability table used when the probability table is switched based on this signal and the data is encoded. Data can be extracted by performing data analysis using a table.

  Therefore, the encoding device and the decoding device have a common probability table in advance, and it is necessary to switch the probability table by a common mechanism.

  In the probability switching signal generator 510, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 510 sets the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, and G inputted via the line L510, and the line L504. The reference frame number of the target block that is input via is acquired (step S2401).

  Note that the motion vector acquired in step S2401 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the motion vector of the encoded block does not exist in step S2401, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S2402).

  Next, it is determined which combination of the five differences calculated in step S2402 has the smallest difference. (Step S2403).

  As a result of the determination in step S2403, if the combination that gives the smallest difference is block B and block F, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block B. A signal Signal_i = 0 (i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as a difference value between the motion vectors of the block B and the block F. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D. When the combination that gives the smallest difference is the block E and the block F, the signal Signal_i = 3 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block G. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block E and block F. When the combination that gives the smallest difference is the block E and the block G, the signal Signal_i = 4 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is used as the difference value between the motion vectors of block E and block G. (Step S2404).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S2404 are output (step S2405), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y).

  Note that the series of processing shown in the flowchart of FIG. 43 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  Also, the difference calculated in step S2402 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. Also in this case, the smallest combination is determined in step S2403. In step S2404, if the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D.

  Further, the difference calculated in step S2402 may be two types, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. Also in this case, the smallest combination is determined in step S2403. In step S2404, if the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G.

  In addition, in order to improve the prediction accuracy when the region including the target block has moved greatly, the positional relationship of the blocks shown in FIG. 41 may be determined as follows.

  The target block is H, and the frame F1201 is a frame including the block H. A frame F1202 is an already encoded frame. The frame F1203 is an already encoded frame.

  In frame F1201, block E is a block including the pixel immediately above the upper leftmost pixel of block H, block F is a block including the pixel immediately above the uppermost left pixel of block H, and block G is the uppermost left of block H. A block including a pixel immediately to the left of the pixel.

  In the frame F1202, the block D is a block that is referenced by the motion vector of the block E, a block that is referenced by the motion vector of the block F, or a block that is referenced by the motion vector of the block G. Also in this case, the block B is a block including a pixel immediately above the upper leftmost pixel of the block D, and the block C is a block including a pixel immediately left of the uppermost left pixel of the block D.

  In the frame F1203, the block A is a block that is referenced by the motion vector of the block D.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block E or the block F refers to It is possible that the block referenced by block or block G or the block referenced by block D may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of these multiple blocks, or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (E, F, G, or D).

  At this time, the data analyzer 501 switches a table of occurrence probability distribution of each index indicating a plurality of motion vector prediction values possessed in advance based on a probability switching signal input via the line L512. Specifically, the index with the highest probability of occurrence is calculated by Signal_i (i = x, y), the probability of occurrence is predicted by MViD (i = x, y), and switching to the probability table is performed accordingly.

  Further, instead of generating a motion vector prediction value in the motion predictor 506, one may be selected based on an index acquired from a plurality of motion vector prediction value determination methods. Details of this method will be described with reference to FIGS. 44 and 45. FIG.

  FIG. 44 is a schematic diagram for explaining a procedure for selecting one based on an index acquired from a plurality of motion vector prediction value determination methods in the motion predictor 506.

  In FIG. 44, the target block is N, and the frame F1301 is a frame including the block N. The frame F1302 is a frame that has already been decoded, and is adjacent to the frame F1301 in the time direction.

  In frame F1301, block J is a block including the pixel immediately above the upper leftmost pixel of block N, block K is a block including the pixel immediately above the uppermost left pixel of block N, and block L is the uppermost right of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In the frame F1302, the block E is a block including a pixel at the same spatial position as the upper left pixel of the block N. Block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, block C is the true of the upper right pixel of block E A block including the upper right pixel, a block D including the pixel immediately left of the lower leftmost pixel of the block E, a block F including a pixel right of the upper rightmost pixel of the block E, and a block G including the block E A block including the pixel directly below the leftmost pixel of block E, a block including the pixel immediately below the pixel below rightmost of block E, and a block I including the pixel directly below the rightmost pixel of block E A block containing

  FIG. 45 is a flowchart for explaining a procedure for selecting one based on an index acquired from a plurality of motion vector prediction value determination methods in the motion predictor 506.

  It is assumed that a prediction value determination method is independently selected for each of the motion vector x component and y component. Specifically, first, a predicted value determination method for the x component of the motion vector is selected and output, and then a predicted value determination method for the y component is selected and output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the motion predictor 506 acquires an index indicating a motion vector prediction value determination method for the target block via the line L514. (Step S2501).

  Next, the index acquired in step S2501 is determined (step S2502).

  As a result of the determination in step S2502, if the index is 0, the past encoded blocks A, B, C, D, E, F, G, H, and I input via the line L510 are stored. When the motion vector average value is the motion vector prediction value and the index is 1, the motion vector of the previously encoded blocks B, D, E, F, and H input via the line L510 If the average value is the motion vector prediction value and the index is 2, the average value of the motion vectors of the previously encoded blocks J, K, L, M, and E that are input via the line L510 is used. A motion vector prediction value is set (step S2503).

  Note that the motion vector acquired in step S2503 is preferably scaled based on the temporal distance between the frame including the target block and the frame referred to by the motion vector of the target block. Since the time of the frame referenced by the motion vector of each block is used for scaling, the reference frame number corresponding to the previously encoded block input via the line L510 and the line L504 are used. It is necessary to obtain the time from the reference frame number of the input target block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  Finally, the motion vector prediction value PMVi (i = x, y) determined in step S2503 is output (step S2504), and the process ends.

  Note that the series of processing shown in the flowchart of FIG. 45 may be determined simultaneously using an index common to the x component and the y component without being divided into the x component and the y component.

  Furthermore, in order to increase the prediction accuracy in the case where the region including the target block has moved greatly, the positional relationship of the blocks shown in FIG. 44 may be determined as follows.

  The target block is N, and the frame F1301 is a frame including the block N. Frame F1302 is a frame that has already been decoded.

  In frame F1301, block J is a block including the pixel immediately above the upper leftmost pixel of block N, block K is a block including the pixel immediately above the uppermost left pixel of block N, and block L is the uppermost right of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In frame F1302, block E is a block referenced by a motion vector of block J, a block referenced by a motion vector of block K, a block referenced by a motion vector of block L, or a motion of block M Let it be the block that the vector references. At that time, block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, and block C is an upper right pixel of block E. A block including the pixel right above the pixel, a block D including the pixel immediately left of the pixel at the bottom left of the block E, a block F including a pixel right above the pixel at the top right of the block E, a block G Is the block containing the pixel directly below the leftmost pixel of block E, block H is the block containing the pixel directly below the rightmost pixel of block E, block I is the block right below the rightmost pixel of block E A block including the lower pixel.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block J or the block K refers to It is possible that the block referenced by block or block L or the block referenced by block M may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (J, K, L, or M).

  Details of the probability switching signal generator 510 will be described with reference to the flowchart of FIG. The probability switching signal indicates the likelihood of the occurrence probability of the index indicating the motion vector prediction value determination method, and switches the probability table based on this signal, and the probability table used when the data is encoded Data can be extracted by performing data analysis using the same probability table.

  Therefore, the encoding device and the decoding device have a common probability table in advance, and it is necessary to switch the probability table by a common mechanism.

  In the probability switching signal generator 510, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 510 moves the decoded blocks A, B, C, D, E, F, G, H, I, J, K, L, and M inputted via the line L510. The reference frame number corresponding to the vector and the reference frame number of the target block input via the line L504 are acquired (step S2601).

  Note that the motion vector acquired in step S2601 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, I, J, K, and L, the magnitudes of motion vectors can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S2601, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Next, there are 4 types of differences: motion vector difference between block A and block E, motion vector difference between block C and block E, motion vector difference between block G and block E, and motion vector difference between block I and block E. The average value of the differences is calculated. In addition, the difference between the motion vector of block B and block E, the difference of the motion vector of block D and block E, the difference of the motion vector of block F and block E, and the difference of the motion vector of block H and block E are 4 ways in total. Calculate the average value of the differences. In addition, the difference between the motion vector of block J and block K, the difference of the motion vector of block J and block M, the difference of the motion vector of block E and block K, and the difference of the motion vector of block E and block M are 4 ways in total. An average value of the differences is calculated (step S2602).

  Subsequently, among the three difference average values calculated in step S2602, the combination having the smallest value is determined. (Step S2603).

  As a result of the determination in step S2603, when the combination that gives the smallest difference average value is block A and E, block C and E, block G and E, and block I and E, the motion vector prediction value of block N is block A signal Signal_i = (i = x, y) 0 indicating that it is likely to be determined as an average value of motion vectors of A, B, C, D, E, F, G, H, and I, and The average difference value MViD_mean (i = x, y) at the time is the difference between the motion vectors of block A and block E, the difference between the motion vectors of block C and block E, the difference of the motion vectors of block G and block E, and the block I and It is set as the average value of a total of four types of differences of the motion vector difference of the block E. When the combination that gives the smallest difference average value is block B and E, block D and E, block F and E, block H and E, the motion vector prediction value of block N is block B, D, E, F, A signal Signal_i = 1 (i = x, y) indicating that it is likely to be determined as an average value of the motion vector of H is generated, and the average difference value MViD_mean (i = x, y) at that time is generated in the block B Difference of motion vectors between block E and block E, difference between motion vectors between block D and block E, difference between motion vectors between block F and block E, and difference between motion vectors between block H and block E. And When the combination that gives the smallest difference average value is block J and K, block J and M, block E and K, block E and M, the motion vector prediction value of block N is block J, K, L, M, A signal Signal_i = 2 (i = x, y) indicating that it is likely to be determined as an average value of the motion vector of E is generated, and the average difference value MViD_mean (i = x, y) at that time is generated in the block J Difference between motion vectors of block K and block K, difference of motion vectors of block J and block M, difference of motion vectors of block E and block K, and difference of motion vectors of block E and block M (Step S2604).

  Finally, the index Index_i (i = x, y) and the motion vector average difference value MViD_mean (i = x, y) determined in step S2604 are output (step S2605), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y).

  Note that the series of processing shown in the flowchart of FIG. 33 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  In step S2602, the average difference value is used as information representing the similarity of the motion vectors of each block group. Instead, the variance value or standard deviation of each block group may be used. In that case, the average difference value generated in step S2604 becomes a variance value or a standard deviation.

  In addition, in order to increase the prediction accuracy when the region including the target block has moved greatly, the positional relationship of the blocks shown in FIG. 44 may be determined as follows.

  The target block is N, and the frame F1301 is a frame including the block N. Frame F1302 is a frame that has already been decoded.

  In frame F1301, block J is a block including the pixel immediately above the upper leftmost pixel of block N, block K is a block including the pixel immediately above the uppermost left pixel of block N, and block L is the uppermost right of block N. A block including the pixel immediately above the rightmost pixel and the block M is a block including a pixel immediately left of the uppermost leftmost pixel of the block N.

  In frame F1302, block E is a block referenced by a motion vector of block J, a block referenced by a motion vector of block K, a block referenced by a motion vector of block L, or a motion of block M Let it be the block that the vector references. At that time, block A is a block including the pixel immediately above the upper left pixel of block E, block B is a block including the pixel immediately above the upper left pixel of block E, and block C is an upper right pixel of block E. A block including the pixel right above the pixel, a block D including the pixel immediately left of the pixel at the bottom left of the block E, a block F including a pixel right above the pixel at the top right of the block E, a block G Is the block containing the pixel directly below the leftmost pixel of block E, block H is the block containing the pixel directly below the rightmost pixel of block E, block I is the block right below the rightmost pixel of block E A block including the lower pixel.

  When the above method is used, when the position of a similar image signal pattern is detected with an accuracy less than the size of each block and the motion vector is detected, the block referred to by the block J or the block K refers to It is possible that the block referenced by block or block L or the block referenced by block M may span multiple blocks with motion vectors. In this case, the average value of the motion vectors of the multiple blocks or the region that spans the blocks It is assumed that the motion vector of the block having the highest ratio is used as the motion vector of the block referred to by the corresponding block (J, K, L, or M).

  At this time, the data analyzer 501 switches a table of occurrence probability distribution of each index indicating a plurality of motion vector prediction values possessed in advance based on a probability switching signal input via the line L512. Specifically, the index with the highest probability of occurrence is calculated by Signal_i (i = x, y), the probability of occurrence is predicted by MViD (i = x, y), and switching to the probability table is performed accordingly.

  Finally, a moving picture decoding program that causes a computer to operate as the moving picture decoding apparatus 50 will be described. FIG. 55 shows a configuration of a moving picture decoding program P50 according to the present embodiment, and the moving picture decoding program P50 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. Note that the configuration of the computer for executing the program stored in the recording medium is the same as the configuration of FIGS. 63 and 64 already described, and therefore, a duplicate description is omitted.

  As shown in FIG. 55, the moving picture decoding program P50 includes a main module P501 that supervises processing, a data analysis module P502, an inverse quantization module P503, an inverse transform module P504, a storage module P505, and a prediction signal. A generation module P506, a motion prediction module P507, an addition module P508, and a probability switching signal generation module P509 are provided.

  The functions realized by the computer by the data analysis module P502, the inverse quantization module P503, the inverse transform module P504, the prediction signal generation module P506, the motion prediction module P507, and the probability switching signal generation module P509 are the data analyzer 501 in FIG. The functions of the inverse quantizer 502, the inverse transformer 503, the prediction signal generator 508, the motion predictor 506, and the probability switching signal generator 510 are the same. The functions that the storage module P505 realizes in the computer are the same as the functions of the frame memory 504 and the motion vector memory 505. The function that the adder module P508 realizes in the computer is the same as the function of the adders 507 and 509.

[Sixth Embodiment]
The moving picture decoding apparatus 60 according to the present invention will be described with reference to FIG.

  As shown in FIG. 47, the moving picture decoding apparatus 60 includes a data analyzer 601, an inverse quantizer 602, an inverse transformer 603, a frame memory 604, and a motion vector memory 605 as functional components. A motion predictor 606, an adder 607, a prediction signal generator 608, an adder 609, and a probability switching signal generator 610.

  The moving picture decoding apparatus 60 receives the compressed stream encoded by the moving picture encoding apparatus described above. This compressed stream includes a residual signal that represents the difference between the original signal of the block to be decoded and the prediction signal, and information related to the generation of the prediction signal of the block to be decoded.

  The data analyzer 601 analyzes the compressed stream input via the line L601 to extract a residual signal of a block (hereinafter referred to as a target block) to be decoded and subjected to orthogonal transform and quantization. Output via line L602. Further, the reference frame number of the target block is extracted and output via the line L604. In addition, based on the probability switching signal input via the line L612, the probability table used for data extraction is switched by the method described later, and then the motion vector difference that is the difference between the motion vector of the target block and its predicted value. The value is extracted and output via line L603.

  The inverse quantizer 602 inversely quantizes the quantized transform coefficient input via the line L602 and outputs the restored transform coefficient via the line L605.

  The inverse transformer 603 performs inverse orthogonal transform on the transform coefficient input via the line L605, and outputs a restored residual signal via the line L606.

  The frame memory 604 stores the image signal of the restored target block input via the line L607 together with a frame image signal (hereinafter referred to as a reference frame image signal) that has been decoded in the past, and if necessary, And output via line L608.

  The motion vector memory 605 includes the motion vector of the target block input via the line L609 and the target input via the line L604 together with the previously decoded motion vector and the corresponding reference frame number. The block reference frame number is stored and output via line L610 as necessary.

  The motion predictor 606 inputs the motion vector of the previously encoded block input via the line L610, the reference frame number corresponding thereto, and the reference frame number of the target block input via the line L604. Then, the motion vector prediction value of the target block is calculated and output via the line L611. The detailed operation of the motion predictor 606 is assumed to be the same as that of the motion predictor 406, and the description thereof is omitted.

  The adder 607 adds the motion vector prediction value input via the line L611 and the motion vector difference value input via the line L603, restores the motion vector of the target block, and passes the line L609. And output.

  The prediction signal generator 608 is input via the line L608, the motion vector signal value of the target block input via the line L609, the reference frame number of the target block input via the line L604, and the line L608. From the reference frame image signal, a predicted image signal for the target block is generated and output via the line L613.

  The adder 609 adds the residual signal input via the line L606 and the prediction image signal for the target block input via the line L613, restores the image signal of the target block, and sets the line L607 Output via.

  The probability switching signal generator 610 inputs the motion vector of the previously encoded block input via the line L610, the corresponding reference frame number, and the reference frame of the target block input via the line L604. From the number, a probability switching signal indicating a probability table to be switched according to a flow to be described later is generated and output via a line L612.

  It is preferable that the detailed operation of the probability switching signal generator 610 corresponds to the operation of the motion predictor 606. In the following, the motion vector prediction value is determined from the motion vector of the encoded block and will be described in the form of not restoring from the index for indicating the motion vector prediction value. However, the index for indicating the motion vector prediction value is described. The present invention can also be applied to the form of restoring from the above.

  When the operation of the motion predictor 606 is the same as the operation described above with reference to FIGS. 27 and 28 of the motion predictor 406, the detailed operation of the probability switching signal generator 610 is as shown in the schematic diagram of FIG. This will be described with reference to the flowchart of FIG. The description of FIG. 27 is the same as that of the motion predictor 406 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value, and is the same as the probability table used when the probability table is switched based on this signal and the data is encoded. Data of motion vector difference values can be extracted by performing data analysis using the probability table.

  Therefore, the encoding device and the decoding device have a common probability table in advance, and it is necessary to switch the probability table by a common mechanism.

  In the probability switching signal generator 610, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 610 passes through the line L604 and the reference frame number corresponding to the motion vectors of the decoded blocks A, B, C, D, E, and F inputted via the line L610. The reference frame number of the target block input in step S2701 is acquired.

  Note that the motion vector acquired in step S2701 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, and F motion vector magnitudes can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S2701, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block E, the difference of the motion vector of block C and the motion vector of block F, the difference of the motion vector of block D and the motion vector of block E, the motion vector of block D And four different motion vector differences of block F are calculated (step S2702).

  Next, it is determined which combination of the four differences calculated in step S2702 has the smallest difference. (Step S2703).

  As a result of the determination in step S2703, if the combination that gives the smallest difference is block B and block E, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block A. A signal Signal_i (i = x, y) = 0 indicating the difference between the motion vectors of the block B and the block E is used as a difference MViD (i = x, y) at that time. When the combination that gives the smallest difference is the block C and the block F, the signal Signal_i = 0 (which indicates that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block A. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block F. When the combination that gives the smallest difference is the block D and the block E, the signal Signal_i = 1 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block E. When the combination that gives the smallest difference is the block D and the block F, the signal Signal_i = 2 () indicating that the motion vector prediction value of the block H is likely to be determined to be the same value as the motion vector of the block E. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is set as the difference value between the motion vectors of block D and block F. (Step S2704).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S2704 are output (step S2705), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Further, only MViD (i = x, y) may be used.

  The series of processes shown in the flowchart of FIG. 48 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and generates a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the data analyzer 601 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on the probability switching signal input via the line L612. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  When the operation of the motion predictor 606 is the same as the operation described above with reference to FIGS. 31 and 32 of the motion predictor 406, the detailed operation of the probability switching signal generator 610 is as shown in the schematic diagram of FIG. This will be described with reference to the flowchart of FIG. The description of FIG. 31 is the same as that of the motion predictor 406 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value, and is the same as the probability table used when the probability table is switched based on this signal and the data is encoded. Data of motion vector difference values can be extracted by performing data analysis using the probability table.

  Therefore, the encoding device and the decoding device have a common probability table in advance, and it is necessary to switch the probability table by a common mechanism.

  In the probability switching signal generator 610, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 610 sets the reference frame number corresponding to the motion vector of the encoded blocks A, B, C, D, E, F, and G input via the line L610, and the line L604. The reference frame number of the target block that is input via is acquired (step S2801).

  Note that the motion vector acquired in step S2801 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  Encoded blocks A and B obtained by scaling based on the time distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block , C, D, E, F, and G motion vector magnitudes can be corrected. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the encoded block itself or the motion vector of the encoded block does not exist in step S2801, the block is regarded as having a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

  Subsequently, the difference between the motion vector of block B and the motion vector of block F, the difference of the motion vector of block C and the motion vector of block G, the difference of the motion vector of block A and the motion vector of block D, the motion vector of block E And the difference between the motion vector of block F and the difference between the motion vector of block E and the motion vector of block G are calculated (step S2802).

  Next, it is determined which combination of the five differences calculated in step S2802 has the smallest difference. (Step S2803).

  As a result of the determination in step S2803, if the combination that gives the smallest difference is block B and block F, it is likely that the motion vector prediction value of block H is determined to be the same value as the motion vector of block B. A signal Signal_i = 0 (i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as a difference value between the motion vectors of the block B and the block F. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D. When the combination that gives the smallest difference is the block E and the block F, the signal Signal_i = 3 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block G. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block E and block F. When the combination that gives the smallest difference is the block E and the block G, the signal Signal_i = 4 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block F. i = x, y) is generated, and the motion vector difference value MViD (i = x, y) at that time is used as the difference value between the motion vectors of block E and block G. (Step S2804).

  Finally, the signal values Signal_i (i = x, y) and MViD (i = x, y) determined in step S2804 are output (step S705), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Further, only MViD (i = x, y) may be used.

  The series of processes shown in the flowchart of FIG. 49 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and generates a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  The difference calculated in step S2802 is the difference between the motion vector of block B and the motion vector of block F, the difference between the motion vector of block C and the motion vector of block G, the motion vector of block A and the motion vector of block D. Three types of differences may be used. At that time, the smallest combination is determined in step S2803. In step S2804, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G. When the combination that gives the smallest difference is the block A and the block D, the signal Signal_i = 2 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block D. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block A and block D.

  Also, the difference calculated in step S2802 may be two types, that is, the difference between the motion vector of block B and the motion vector of block F, and the difference between the motion vector of block C and the motion vector of block G. At that time, the smallest combination is determined in step S2803. In step S2804, when the combination that gives the smallest difference is block B and block F, the motion vector prediction value of block H is the same as the motion vector of block B. A signal Signal_i = 0 (i = x, y) indicating that it is plausible to be determined as a value is generated, and the difference MViD (i = x, y) at that time is calculated as the motion vector of the block B and the block F. The difference value. When the combination that gives the smallest difference is the block C and the block G, the signal Signal_i = 1 () indicating that it is likely that the motion vector prediction value of the block H is determined to be the same value as the motion vector of the block C. i = x, y) is generated, and the difference MViD (i = x, y) at that time is set as the difference value between the motion vectors of block C and block G.

  The series of processes shown in the flowchart of FIG. 49 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component, and generates a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  At this time, the data analyzer 601 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on the probability switching signal input via the line L612. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  When the operation of the motion predictor 606 is the same as the operation of the motion predictor 406 described with reference to FIGS. 33 and 34, the detailed operation of the probability switching signal generator 610 is shown in the schematic diagram of FIG. This will be described with reference to 50 flowcharts. The description of FIG. 33 is the same as that in the motion predictor 406 described above.

  At this time, the probability switching signal indicates the likelihood of the occurrence probability distribution of the motion vector difference value, and is the same as the probability table used when the probability table is switched based on this signal and the data is encoded. Data of motion vector difference values can be extracted by performing data analysis using the probability table.

  In the probability switching signal generator 610, the x component and y component of the motion vector are calculated independently. Specifically, the x component of the motion vector is first calculated, a probability switching signal corresponding to the x component is output, then the y component is calculated, and a probability switching signal corresponding to the y component is output. In the following, a subscript indicating a component is written as i (i is x or y) as necessary.

  First, the probability switching signal generator 610 includes the reference frame numbers corresponding to the motion vectors of the decoded blocks A, B, C, D, E, F, G, H, and I input via the line L610. The reference frame number of the target block input via the line L604 is acquired (step S2901).

  Note that the motion vector acquired in step S2901 is preferably scaled based on a temporal distance between a frame including the target block and a frame referred to by the motion vector of the target block. Since the time of the frame referred to by the motion vector of each block is used at the time of scaling, it is necessary to obtain the time from the reference frame number corresponding to each block.

  The obtained decoded blocks A and B based on the temporal distance (t0-te) from the time t0 of the frame including the target block to the time te of the frame referred to by the motion vector of the target block by scaling. , C, D, E, F, G, H, and I can correct the magnitude of the motion vector. As a specific example, if the time of the frame referenced by the motion vector MViA (i = x, y) of the block A is ta, the scaled motion vector MViA ′ (i = x, y) The vector given by 1).

  Note that this scaling processing is not essential, and reference frame numbers corresponding to the respective blocks are not required when the scaling processing is not performed.

  If the decoded block itself or the motion vector of the decoded block does not exist in step S2901, it is assumed that the block has a very large motion vector, and the process proceeds with a sufficiently large value. Alternatively, it is excluded from the processing target.

Subsequently, variance of motion vector of block A, motion vector of block B and motion vector of block C, variance of motion vector of block D, motion vector of block E and motion vector of block F, motion vector of block G and block Three sets of variances of H motion vector and block I motion vector variance are calculated. (Step S2902)
Next, it is determined which combination of the three variances calculated in step S2902 has the smallest variance. (Step S2903).

  As a result of the determination in step S2903, when the combination that gives the smallest difference is block A, block B, and block C, the motion vector prediction value of block J is the motion vector of block A, the motion vector of block B, and the block A signal Signal_i = 0 (i = x, y) indicating that it is likely to be determined as an average value of the motion vector of C is generated, and the variance MVi_var (i = x, y) at that time is generated as a block A It is assumed that the motion vectors of block B and block C are distributed. When the combination that gives the smallest difference is block D, block E, and block F, the motion vector prediction value of block J is the average of the motion vector of block D, the motion vector of block E, and the motion vector of block F. A signal Signal_i = 1 (i = x, y) indicating that it is likely to be determined is generated, and a distribution MVi_var (i = x, y) at that time is used as a motion vector of the block D, the block E, and the block F. Of variance. When the combination that gives the smallest difference is block G, block H, and block I, the motion vector prediction value of block J is the average value of the motion vector of block G, the motion vector of block H, and the motion vector of block I. A signal Signal_i = 2 (i = x, y) indicating that it is likely to be determined is generated, and a distribution MVi_var (i = x, y) at that time is used as a motion vector of the block G, the block H, and the block I Of variance. (Step S2904).

  Finally, the signal values Signal_i (i = x, y) and MVi_var (i = x, y) determined in step S2904 are output (step S2905), and the process ends.

  The occurrence probability signal may be only Signal_i (i = x, y). Further, only MVi_var (i = x, y) may be used.

  The series of processes shown in the flowchart of FIG. 50 calculates a vector difference (specifically, a vector norm) without dividing the x component and the y component into a probability switching signal corresponding to the x component and the y component. It may be determined at the same time.

  Further, although variance is used as a value representing the similarity of motion vectors in step S2902, a standard deviation may be used. Further, a motion vector difference value between blocks or an average of the plurality of difference values may be used.

  In step S2904, a signal indicating that it is likely that the average value of the three motion vectors is used as the predicted value of the motion vector is generated. The intermediate value of any one of the three motion vectors or any one of the values is used as the predicted value. It may indicate that it is plausible to do.

  In FIG. 50, the probability switching signal is determined with reference to three blocks of motion vectors that are continuous in the time direction and the space direction, but only the space direction may be used. Further, only two blocks or four or more blocks may be used in each direction.

  At this time, the data analyzer 601 switches a plurality of motion vector difference value occurrence probability distribution tables possessed in advance based on a probability switching signal input via the line L612. Specifically, the direction (x-axis direction, y-axis direction, or temporal direction) for quoting the motion vector prediction value is predicted by Signal_i (i = x, y), and the motion vector difference value is calculated by MViD_var (i = x, y). The occurrence probability is predicted, and the probability table is switched accordingly.

  Finally, a video decoding program for causing a computer to operate as the video decoding device 60 will be described. FIG. 56 shows a configuration of a moving picture decoding program P60 according to the present embodiment, and the moving picture decoding program P60 is provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, and a ROM, or a semiconductor memory. Note that the configuration of the computer for executing the program stored in the recording medium is the same as the configuration of FIGS. 63 and 64 already described, and therefore, a duplicate description is omitted.

  As shown in FIG. 56, the moving picture decoding program P60 includes a main module P601 that supervises processing, a data analysis module P602, an inverse quantization module P603, an inverse transform module P604, a storage module P605, and a prediction signal. A generation module P606, a motion prediction module P607, an addition module P608, and a probability switching signal generation module P609 are provided.

  The functions realized by the data analysis module P602, inverse quantization module P603, inverse transform module P604, prediction signal generation module P606, motion prediction module P607, and probability switching signal generation module P609 are the data analyzer 601 in FIG. The functions of the inverse quantizer 602, the inverse transformer 603, the prediction signal generator 608, the motion predictor 606, and the probability switching signal generator 610 are the same. The functions that the storage module P605 realizes in the computer are the same as the functions of the frame memory 604 and the motion vector memory 605. The function that the adder module P608 realizes in the computer is the same as the function of the adders 607 and 609.

It is the schematic which shows an example of the moving image encoder concerning 1st Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 1st Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 1st Embodiment. It is the schematic which shows an example of the moving image encoder concerning 2nd Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 2nd Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 2nd Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 2nd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 2nd Embodiment. It is the schematic which shows an example of the moving image encoder concerning 3rd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 3rd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 3rd Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 3rd Embodiment. It is the schematic which shows an example of the moving image decoding apparatus concerning 4th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 4th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 4th Embodiment. It is the schematic which shows an example of the moving image decoding apparatus concerning 5th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 5th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 5th Embodiment. It is a schematic diagram used for description of operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the motion predictor in 5th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 5th Embodiment. It is the schematic which shows an example of the moving image decoding apparatus concerning 6th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 6th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 6th Embodiment. It is a flowchart which shows operation | movement of the probability switching signal generator in 6th Embodiment. It is a module figure of the moving image encoding program concerning 1st Embodiment. It is a module diagram of the moving image encoding program concerning 2nd Embodiment. It is a module diagram of the moving image encoding program concerning 3rd Embodiment. It is a module diagram of the moving image decoding program concerning 4th Embodiment. It is a module diagram of the moving image decoding program concerning 5th Embodiment. It is a module figure of the moving image decoding program concerning 6th Embodiment. 3 is a flowchart illustrating an example in which a peripheral motion vector utilization method based on weighting is applied to the block of FIG. 2 in the first embodiment. 7 is a flowchart illustrating an example in which a peripheral motion vector utilization method based on weighting is applied to the block of FIG. 6 in the first embodiment. 9 is a flowchart illustrating an example in which a peripheral motion vector utilization method based on weighting is applied to the block of FIG. 8 in the first embodiment. It is a flowchart which shows the example which applied the surrounding motion vector utilization method by weighting to the block of FIG. 27 in 4th Embodiment. FIG. 32 is a flowchart illustrating an example in which a peripheral motion vector utilization method based on weighting is applied to the block of FIG. 31 in the fourth embodiment. It is a flowchart which shows the example which applied the surrounding motion vector utilization method by weighting to the block of FIG. 33 in 4th Embodiment. It is a figure which shows the hardware constitutions of the computer for performing the program recorded on the recording medium. It is a perspective view of a computer for executing a program stored in a recording medium.

Explanation of symbols

  10, 20, 30 ... moving picture encoding device, 40, 50, 60 ... moving picture decoding device, P10, P20, P30 ... moving picture encoding program, P40, P50, P60 ... moving picture decoding program, 60 ... Recording medium 62 ... reading device 64 ... working memory 66 ... memory 68 ... display 70 ... mouse 72 ... keyboard 74 ... communication device 76 ... CPU 80 ... computer 90 ... computer data signal

Claims (58)

Region dividing means for dividing an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, motion detection means for detecting a motion vector of the processing target region, and a prediction signal for the processing target region Predictive signal generating means for generating the residual signal, residual signal generating means for generating a residual signal between the predicted signal and the processing target area signal, and generating the compressed data by encoding the residual signal and the motion vector Encoding means,
The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. A moving picture coding apparatus characterized by coding the motion vector based on the above. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The motion image encoding apparatus according to claim 1, wherein a motion vector prediction value is determined based on the motion vector prediction value. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The motion picture encoding apparatus according to claim 2, wherein a motion vector prediction value determination method is selected based on the motion vector prediction value determination method. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The moving picture coding apparatus according to claim 2, wherein a predicted value of the motion vector is determined by weighting based on the weighting. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The video encoding apparatus according to claim 1, wherein a probability distribution used when encoding information for indicating a motion vector prediction value is adaptively determined. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. 6. The moving picture encoding apparatus according to claim 5, wherein a probability distribution used when encoding information for indicating a motion vector prediction value determination method is adaptively determined. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be encoded. The video encoding apparatus according to claim 1, wherein a probability distribution used when encoding the motion vector difference value is adaptively determined based on. Region dividing means for dividing an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, motion detection means for detecting a motion vector of the processing target region, and a prediction signal for the processing target region Predictive signal generating means for generating the residual signal, residual signal generating means for generating a residual signal between the predicted signal and the processing target area signal, and generating the compressed data by encoding the residual signal and the motion vector Encoding means,
The prediction signal generation unit encodes the motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be encoded. A video encoding apparatus characterized by the above. The prediction signal generation means determines a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be encoded. The moving picture coding apparatus according to claim 8, wherein: The prediction signal generation means is a method for determining a prediction value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be encoded The moving picture coding apparatus according to claim 9, wherein the moving picture coding apparatus is selected. The prediction signal generation means calculates a predicted value of a motion vector by weighting based on similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. The moving picture coding apparatus according to claim 9, wherein the moving picture coding apparatus is determined. The prediction signal generation means indicates a predicted value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. 9. The moving picture encoding apparatus according to claim 8, wherein a probability distribution used when encoding information for encoding is adaptively determined. The prediction signal generation means is a method for determining a prediction value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be encoded The moving picture encoding apparatus according to claim 12, wherein a probability distribution used when encoding information for indicating is adaptively determined. The prediction signal generating means encodes a motion vector difference value based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. 9. The moving picture encoding apparatus according to claim 8, wherein a probability distribution used for the determination is adaptively determined. Data extraction means for extracting residual signal and motion vector information relating to the processing target area from the compressed data, residual signal restoration means for restoring the residual signal relating to the processing target area to a reproduction residual signal, and the processing A prediction signal generation unit that generates a prediction signal for the target region; and an image restoration unit that restores the pixel signal of the target region by adding the prediction signal and the reproduction residual signal,
The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. A motion picture decoding apparatus characterized in that the motion vector is decoded based on the motion vector. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The motion picture decoding apparatus according to claim 15, wherein a motion vector prediction value is determined based on the motion vector prediction value. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The motion picture decoding apparatus according to claim 16, wherein a motion vector prediction value determination method is selected based on the motion vector prediction value determination method. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The moving picture decoding apparatus according to claim 16, wherein a predicted value of a motion vector is determined by weighting based on the weighting. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The video decoding device according to claim 15, wherein a probability distribution used when decoding information for indicating a predicted value of a motion vector is adaptively determined. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. 20. The moving picture decoding apparatus according to claim 19, wherein a probability distribution used when decoding information for indicating a motion vector prediction value determination method is adaptively determined. The prediction signal generation means may determine the similarity of motion vectors of a plurality of blocks in different frames, or the similarity of motion vectors of a plurality of blocks in the same frame as the block to be decoded. The video decoding device according to claim 15, wherein the probability distribution used when decoding the motion vector difference value is adaptively determined based on the probability distribution. Data extraction means for extracting residual signal and motion vector information relating to the processing target area from the compressed data, residual signal restoration means for restoring the residual signal relating to the processing target area to a reproduction residual signal, and the processing A prediction signal generation unit that generates a prediction signal for the target region; and an image restoration unit that restores the pixel signal of the target region by adding the prediction signal and the reproduction residual signal,
The prediction signal generating means decodes the motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded. A moving picture decoding apparatus characterized by the above. The prediction signal generation means determines a motion vector prediction value based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be decoded The moving picture decoding apparatus according to claim 22, wherein the moving picture decoding apparatus is provided. The prediction signal generation means is a method for determining a prediction value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be decoded The moving picture decoding apparatus according to claim 23, wherein: The predicted signal generation means calculates a predicted value of a motion vector by weighting based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be decoded. The moving picture decoding apparatus according to claim 23, wherein the moving picture decoding apparatus is determined. The prediction signal generation means indicates a predicted value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be decoded 23. The moving picture decoding apparatus according to claim 22, wherein a probability distribution to be used when decoding information for decoding is adaptively determined. The prediction signal generation means is a method for determining a prediction value of a motion vector based on the similarity of motion vectors of a plurality of blocks including a block separated by one or more blocks from the target block in the same frame as the block to be decoded 27. The moving picture decoding apparatus according to claim 26, wherein a probability distribution to be used when decoding information for indicating the information is adaptively determined. The prediction signal generation means decodes a motion vector difference value based on similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded. The moving picture decoding apparatus according to claim 22, wherein the probability distribution to be used in the determination is adaptively determined. A video encoding method executed by a video encoding device,
A region dividing step for dividing an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, a motion detection step for detecting a motion vector of the processing target region, and a prediction signal for the processing target region A prediction signal generation step of generating a residual signal, a residual signal generation step of generating a residual signal between the prediction signal and the processing target region signal, and encoding the residual signal and the motion vector to generate compressed data An encoding step,
In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded A moving image encoding method, wherein the motion vector is encoded based on similarity of motion vectors of blocks. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded 30. The moving picture coding method according to claim 29, wherein a predicted value of the motion vector is determined based on similarity of the motion vectors of the blocks. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded 31. The moving picture coding method according to claim 30, wherein a motion vector prediction value determination method is determined based on similarity between motion vectors of blocks. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded The moving picture coding method according to claim 30, wherein the predicted value of the motion vector is determined by weighting based on the similarity of the motion vectors of the blocks. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded 30. The moving picture code according to claim 29, wherein a probability distribution used when coding information for indicating a predicted value of a motion vector is adaptively determined based on similarity between motion vectors of blocks. Method. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded 34. The moving image according to claim 33, wherein a probability distribution used when encoding information for indicating a method for determining a predicted value of a motion vector is adaptively determined based on similarity between motion vectors of blocks. Image coding method. In the prediction signal generation step, the moving image encoding device is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be encoded 30. The moving picture coding method according to claim 29, wherein the probability distribution used when coding the motion vector difference value is adaptively determined based on the similarity of the motion vectors of the blocks. A video encoding method executed by a video encoding device,
A region dividing step for dividing an input moving image signal composed of a time sequence of frame image signals into a plurality of processing target regions, a motion detection step for detecting a motion vector of the processing target region, and a prediction signal for the processing target region A prediction signal generation step of generating a residual signal, a residual signal generation step of generating a residual signal between the prediction signal and the processing target region signal, and encoding the residual signal and the motion vector to generate compressed data An encoding step,
In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. A moving image encoding method characterized by encoding the motion vector. In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. The motion image encoding method according to claim 36, wherein a predicted value of the motion vector is determined. In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. 38. The moving picture coding method according to claim 37, wherein a motion vector prediction value determination method is determined. In the prediction signal generating step, the video encoding device weights based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be encoded. The motion image encoding method according to claim 37, wherein a predicted value of a motion vector is determined by: In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. 37. The moving picture coding method according to claim 36, wherein a probability distribution used when coding information for indicating a predicted value of a motion vector is adaptively determined. In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. 41. The moving picture encoding method according to claim 40, wherein a probability distribution used when encoding information for indicating a motion vector prediction value determination method is adaptively determined. In the prediction signal generation step, the video encoding device is based on the similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be encoded. The video encoding method according to claim 36, wherein a probability distribution used when encoding the motion vector difference value is adaptively determined. A moving picture decoding method executed by a moving picture decoding apparatus,
A data extraction step for extracting residual signal and motion vector information relating to the processing target region from the compressed data, a residual signal restoring step for restoring the residual signal relating to the processing target region to a reproduction residual signal, and the processing A prediction signal generation step of generating a prediction signal for the target region, and an image restoration step of restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal,
In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded A moving picture decoding method, wherein the motion vector is decoded based on similarity of block motion vectors. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded 44. The moving picture decoding method according to claim 43, wherein a predicted value of the motion vector is determined based on similarity of the motion vectors of the blocks. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded 45. The moving picture decoding method according to claim 44, wherein a motion vector prediction value determination method is determined based on similarity between motion vectors of blocks. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded 45. The moving picture decoding method according to claim 44, wherein the predicted value of the motion vector is determined by weighting based on the similarity of the motion vectors of the blocks. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded 44. The moving picture decoding according to claim 43, wherein a probability distribution used in decoding information for indicating a predicted value of a motion vector is adaptively determined based on similarity between motion vectors of blocks. Method. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded The moving image according to claim 47, wherein a probability distribution used when decoding information for indicating a method for determining a predicted value of a motion vector is adaptively determined based on similarity between motion vectors of blocks. Image decoding method. In the prediction signal generation step, the moving picture decoding apparatus is configured such that the motion vectors of a plurality of blocks in different frames, or a plurality of blocks in the same frame as the similarity and the block to be decoded 44. The moving picture decoding method according to claim 43, wherein the probability distribution used when decoding the motion vector difference value is adaptively determined based on the similarity of the motion vectors of the blocks. A moving picture decoding method executed by a moving picture decoding apparatus,
A data extraction step for extracting residual signal and motion vector information relating to the processing target region from the compressed data, a residual signal restoring step for restoring the residual signal relating to the processing target region to a reproduction residual signal, and the processing A prediction signal generation step of generating a prediction signal for the target region, and an image restoration step of restoring the pixel signal of the target region by adding the prediction signal and the reproduction residual signal,
In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. Decoding the motion vector. A moving image decoding method, wherein: In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. The motion picture decoding method according to claim 50, wherein a predicted value of the motion vector is determined. In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. 52. The moving picture decoding method according to claim 51, further comprising: determining a motion vector prediction value determination method. In the prediction signal generating step, the moving picture decoding apparatus weights based on similarity of motion vectors of a plurality of blocks including a block separated by one block or more from the target block in the same frame as the block to be decoded 52. The moving picture decoding method according to claim 51, wherein a predicted value of a motion vector is determined by: In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. 51. The moving picture decoding method according to claim 50, wherein a probability distribution used when decoding information for indicating a motion vector prediction value is adaptively determined. In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. 55. The moving picture decoding method according to claim 54, wherein a probability distribution used when decoding information for indicating a motion vector prediction value determination method is adaptively determined. In the prediction signal generation step, the video decoding device is configured based on the similarity of motion vectors of a plurality of blocks including a block separated from the target block by one block or more in the same frame as the block to be decoded. 51. The moving picture decoding method according to claim 50, wherein the probability distribution used when decoding the motion vector difference value is adaptively determined.   A moving picture coding program for causing a computer to function as the moving picture coding apparatus according to claim 1.   A moving picture decoding program for causing a computer to function as the moving picture decoding apparatus according to any one of claims 15 to 28.

Images