JP6242517B2 - Moving picture predictive decoding apparatus and moving picture predictive decoding method - Google Patents

Description

  The present invention relates to a moving picture predictive decoding apparatus and method, and more particularly, to filter processing for reference samples used for predictive coding within a screen.

  In order to efficiently transmit and store moving image data, a compression encoding technique is used. In the case of moving images, MPEG1-4 and H.264 are used. 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. In predictive coding within a screen, a predicted signal is generated using an adjacent previously reproduced image signal (reconstructed compressed image data) in the same screen as the target block, and then the predicted signal is The differential signal subtracted from the signal of the target block is encoded. In predictive coding between screens, the adjacent reproduced image signal in the screen different from the target block is referred to, the motion is corrected, the predicted signal is generated, and the predicted signal is subtracted from the signal of the target block. Encode the difference signal.

  In normal inter-screen prediction (inter-prediction) encoding, a prediction signal is generated by a method in which a signal similar to the pixel signal is searched from a previously reproduced screen for a block to be encoded. Then, a motion vector that is a spatial displacement amount between the target block and a region formed by the searched signal, and a residual signal between the pixel signal and the prediction signal of the target block are encoded. Such a method for searching for a motion vector for each block is called block matching.

  FIG. 10 is a schematic diagram for explaining the block matching process. Here, a procedure for generating a prediction signal will be described using the target block 702 on the encoding target screen 701 as an example. The reference screen 703 has already been reproduced, and the area 704 is an area in the same position as the target block 702. In the block matching, a search range 705 surrounding the region 704 is set, and a region 706 in which the absolute value error sum with the pixel signal of the target block 702 is minimum is detected from the pixel signal in this search range. The signal in the area 706 becomes a prediction signal, and the amount of displacement from the area 704 to the area 706 is detected as a motion vector 707. Also, a method of preparing a plurality of reference screens 703, selecting a reference screen for performing block matching for each target block, and detecting reference screen selection information is often used. H. H.264 provides a plurality of prediction types having different block sizes for encoding motion vectors in order to cope with changes in local features of an image. H. H.264 prediction types are described in Patent Document 2, for example.

  H. In the H.264 intra-screen prediction (intra prediction) encoding, a method of generating a prediction signal by extrapolating the already reproduced pixel values adjacent to the encoding target block in a predetermined direction is adopted. FIG. 2 is a schematic diagram for explaining an intra-screen prediction method used for H.264. In FIG. 11A, a target block 802 is a block to be encoded, and a pixel group (reference sample group) 801 including pixels A to M adjacent to the boundary of the target block 802 is an adjacent region. It is an image signal that has already been reproduced in past processing.

  In this case, a pixel group (reference sample group) 801 that is an adjacent pixel immediately above the target block 802 is extended downward to generate a prediction signal. In FIG. 11B, the already reproduced pixels (I to L) on the left of the target block 804 are stretched to the right to generate a prediction signal. A specific method for generating a prediction signal is described in Patent Document 1, for example. Thus, the difference between each of the nine prediction signals generated by the method shown in FIGS. 11A to 11I and the pixel signal of the target block is taken, and the one with the smallest difference value is used as the optimum prediction signal. . As described above, a prediction signal (intra prediction sample) can be generated by extrapolating pixels. The above contents are described in Patent Document 1 below.

  In addition, in the in-screen prediction shown in Non-Patent Document 1, 25 types (34 types in total) of prediction signal generation methods with different reference sample stretching directions are prepared in addition to the above nine types.

  In Non-Patent Document 1, a low-pass filter is applied to a reference sample before generating a prediction signal in order to suppress distortion generated in the reference sample. Specifically, extrapolation prediction is performed after applying a 121 filter with a weighting factor of 1: 2: 1 to the reference samples. This process is called intra smoothing.

  The in-screen prediction of Non-Patent Document 1 will be described with reference to FIGS. FIG. 7 shows an example of block division. Five blocks 220, 230, 240, 250, 260 adjacent to the target block 210 having a block size of N × N samples have already been reproduced. For intra prediction of the target block 210, reference samples indicated by ref [x] (x = 0 to 4N) are used. FIG. 8 shows a processing flow of intra prediction. First, in step 310, the prediction signal generator that performs the intra-screen prediction processing acquires reference samples ref [x] (x = 0 to 4N) from the memory that stores the reproduced pixels. At this time, there are cases where adjacent blocks have not been reproduced yet for reasons such as the coding order, and all 4N + 1 reference samples ref [x] cannot be acquired. In that case, 4N + 1 reference samples are prepared by substituting non-existing samples by padding processing (copying existing sample values nearby). The details of the padding process are described in Non-Patent Document 1. Next, in step 320, the prediction signal generator performs a smoothing process on the reference sample with a 121 filter. Finally, in step 330, the prediction signal generator estimates the signal in the target block by extrapolation (direction of intra-screen prediction), and generates a prediction signal (intra prediction sample).

US Pat. No. 6,765,964 US Patent Publication No. 7003035

B. Bross et. Al, “High efficiency video coding (HEVC) text specification draft 8”, Joint Collaborative Team on Video Coding (JCT-VC) of ITU-T SG16 WP3 and ISO / IEC JTC1 / SC29 / WG11, JCTVC- J1003, 10th Meeting: Stockholm, Sweden, 11-20 July, 2012.

  FIG. 9 shows an example of a flat region signal having similar pixel values. When the original pixel value (original sample value) 410 is encoded by coarse quantization, the reproduction value (reproduction sample value) 420 in the block is constant. Value, and step-like distortion occurs at the block boundary 430. This distortion is known as block noise, and is usually removed by applying a filter that removes block noise to the reproduced image. However, since the reference sample used for the intra prediction is a signal before the filter process for removing the block noise, the block noise remaining in the reference sample at the block boundary is predicted by the prediction signal of the target block ( Propagate to intra prediction sample). Since the block noise that has propagated to the prediction signal cannot be removed by the block noise removal processing for the reproduction signal, it is propagated as it is to the reference sample group of the next target block.

  In Non-Patent Document 1, since 34 different extrapolation directions are prepared for the extrapolation method of intra prediction (direction of intra prediction), block noise propagates while changing the direction. As a result, a plurality of pseudo contours are generated in the reproduction signal of the flat area in the image. In particular, when noise propagates to a large block, the pseudo contour crosses the large block, which has a large visual effect.

  The 121 filter described in the background art has an effect of removing noise in the reference sample, but cannot remove stepped noise as shown in FIG. 9 because the number of taps is short.

  Therefore, an object of the present invention is to suppress artificial noise such as the above-described pseudo contour.

  A video predictive decoding apparatus according to an aspect of the present invention includes an intra prediction mode and a residual that indicate an intra-screen prediction method for a target block to be decoded from compressed data that is encoded by being divided into a plurality of blocks. Decoding means for decoding the compressed data of the signal, prediction signal generating means for generating an intra-screen prediction signal using the intra-prediction mode and the already reproduced reference sample adjacent to the target block, and the residual signal Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data, and the pixel signal of the target block is restored by adding the prediction signal and the reproduction residual signal, and the restored Block storage means for storing the pixel signal of the target block for use as the reference sample, and the prediction signal generation means includes the block storage means. In order to obtain reference samples from the already-reproduced blocks around the target block stored in and to generate an interpolated reference sample, two or more key reference samples at predetermined positions are generated in the reference sample. Interpolating, extrapolating the interpolated reference sample based on the intra prediction mode to generate the intra prediction signal, the key reference sample as three reference samples located at the end and center of the reference sample group, The prediction signal generation means is configured to calculate a difference between an added value of two key reference samples located at one end and the center of the reference sample group and a double value of a reference sample located at the center of the two key reference samples. Of the first absolute value, which is the absolute value of the reference sample, and a predetermined threshold value, and the addition of two key reference samples located at the other end and the center of the reference sample group. A second absolute value, which is an absolute value of a difference between a value and a value twice the reference sample located in the center of the two key reference samples, the result of comparing the threshold value, the size of the target block, Based on the reference sample, the interpolation processing of the reference sample and the smoothing processing of the reference sample are applied and performed, and the prediction signal generation means has the first absolute value smaller than the threshold and the first 2 When the absolute value is smaller than the threshold, interpolation processing of the reference sample is performed, and the threshold is restored by the moving picture predictive decoding apparatus.

  The present invention can also be understood as an invention relating to a video predictive decoding method, and can be described as follows.

  A moving image predictive decoding method according to an aspect of the present invention is a moving image predictive decoding method executed by a moving image predictive decoding device, and decodes from compressed data divided into a plurality of blocks and encoded. A decoding step for decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be processed and compressed data of a residual signal, and the intra-prediction mode and a replayed reference sample adjacent to the target block A prediction signal generation step for generating an intra-screen prediction signal, a residual signal restoration step for restoring the reproduction residual signal of the target block from the compressed data of the residual signal, and the prediction signal and the reproduction residual signal. The pixel signal of the target block is restored by adding, and the restored pixel signal of the target block is stored for use as the reference sample. A block storage step, wherein in the prediction signal generation step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference sample is generated to generate an interpolation reference sample. Interpolating between two or more key reference samples at predetermined positions, generating the intra prediction signal by extrapolating the interpolation reference samples based on the intra prediction mode, refer to the key reference samples Three reference samples located at the end and center of the sample group are used, and in the prediction signal generation step, the sum of two key reference samples located at one end and the center of the reference sample group and the two key reference samples The result of comparing the first absolute value, which is the absolute value of the difference from the double value of the reference sample located in the center of the reference value, with a predetermined threshold value, The second absolute value which is the absolute value of the difference between the sum of the two key reference samples located at the other end and the center of the reference sample group and the double value of the reference sample located at the center of the two key reference samples The prediction signal is obtained by switching between interpolation processing of the reference sample and smoothing processing of the reference sample based on a comparison result between the value and the threshold value, and the size of the target block. In the generating step, when the first absolute value is smaller than the threshold value and the second absolute value is smaller than the threshold value, the reference sample is interpolated, and the threshold value is determined based on the video predictive decoding It is restored by the device.

  According to the filtering process to the reference sample by the bilinear interpolation of the present invention, since the signal in the reference sample is gently changed using the samples at both ends of the reference sample, artificial noise such as a pseudo contour is not generated. Can be suppressed.

It is a block diagram which shows the moving image predictive coding apparatus which concerns on embodiment of this invention. It is a block diagram which shows the moving image predictive decoding apparatus which concerns on embodiment of this invention. It is a flowchart which shows the prediction method in a screen which concerns on embodiment of this invention. It is a flowchart which shows another example of the prediction method in a screen which concerns on embodiment of this invention. It is a figure which shows the hardware constitutions of the computer for performing the program recorded on the recording medium. It is a general-view figure of the computer for performing the program memorize | stored in the recording medium. It is a figure explaining the example of the reference sample used for the prediction in a screen. It is a flowchart which shows the prediction method in a screen in a prior art. It is a figure explaining the relationship between the original signal and reproduction signal in a flat area | region. It is a schematic diagram for demonstrating the motion estimation process in the prediction between screens. It is a schematic diagram for demonstrating the prediction in a screen by extrapolation of a reference sample. It is a figure explaining another example of the reference sample used for the prediction in a screen. It is a flowchart explaining the process of the prediction signal generator 103 of FIG. 3 is a flowchart for explaining processing of a prediction signal generator 208 in FIG. 2. It is a flowchart which shows the 2nd another example of the prediction method in a screen which concerns on embodiment of this invention. It is a block diagram which shows the structure of a moving image predictive encoding program. It is a block diagram which shows the structure of a moving image prediction decoding program.

  Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 7 and FIGS. 13 to 17.

  FIG. 1 is a block diagram showing a video predictive coding apparatus 100 according to an embodiment of the present invention. As shown in FIG. 1, the moving image predictive encoding device 100 includes an input terminal 101, a block divider 102, a prediction signal generator 103, a frame memory 104, a subtractor 105, a converter 106, a quantizer 107, an inverse quantum, and the like. , An adder 110, an entropy encoder 111, an output terminal 112, a block memory 113, and a loop filter 114. The subtractor 105, the converter 106, and the quantizer 107 correspond to “encoding means” recited in the claims. The inverse quantizer 108, the inverse transformer 109, and the adder 110 correspond to “decoding means” recited in the claims. The frame memory 104 corresponds to “image storage means”, and the block memory 113 corresponds to “block storage means”.

  The operation of the moving picture predictive coding apparatus 100 configured as described above will be described below. A moving image signal composed of a plurality of images is input to the input terminal 101. An image to be encoded is divided into a plurality of regions by the block divider 102. In the embodiment according to the present invention, the block size is not limited as illustrated in FIG. A plurality of block sizes and shapes may be mixed on one screen. The coding order of blocks is described in Non-Patent Document 1, for example. Next, a prediction signal is generated for a region to be encoded (hereinafter referred to as a target block). In the embodiment according to the present invention, two types of prediction methods, inter-screen prediction and intra-screen prediction, are used. The prediction signal generation processing in the prediction signal generator 103 will be described later with reference to FIG.

  The subtractor 105 subtracts the prediction signal (via the line L103) from the signal of the target block (via the line L102) to generate a residual signal. The residual signal is subjected to discrete cosine transform by a transformer 106, and each transform coefficient is quantized by a quantizer 107. The entropy encoder 111 encodes the quantized transform coefficient and sends it from the output terminal 112 together with prediction information necessary for generating a prediction signal.

  In order to perform intra prediction or inter prediction for the subsequent target block, the compressed signal of the target block is inversely processed and restored. That is, the quantized transform coefficient is inversely quantized by the inverse quantizer 108 and then inverse discrete cosine transformed by the inverse transformer 109 to restore the residual signal. The residual signal restored by the adder 110 and the prediction signal sent from the line L103 are added to reproduce the signal of the target block. The reproduced block signal is stored in the block memory 113 for intra-screen prediction. A reproduced image constituted by the reproduced signal is stored in the frame memory 104 after block noise generated in the reproduced image is removed by the loop filter unit 114.

  A prediction signal processing flow in the prediction signal generator 103 will be described with reference to FIG. First, in step S302, prediction information necessary for inter-screen prediction is generated. Specifically, using a reproduced image that has been encoded in the past and restored as a reference image, a motion vector that gives a prediction signal with the smallest error with respect to the target block and a reference screen are searched from this reference image. At this time, the target block is input via line L102 and the reference image is input via L104. As the reference image, a plurality of images encoded and restored in the past are used as the reference image. For details, see the conventional technology H.264. H.264 or the method shown in Non-Patent Document 1.

  In step S303, prediction information necessary for intra-screen prediction is generated. As illustrated in FIG. 7, a prediction signal is generated for a plurality of in-screen prediction directions using already reproduced pixel values spatially adjacent to the target block. And the prediction direction (intra prediction mode) which gives the prediction signal with the smallest error with respect to an object block is determined. At this time, the prediction signal generator 103 obtains already reproduced pixel signals in the same screen as reference samples from the block memory 113 via the line L113, and generates an intra-screen prediction signal by extrapolating these signals. To do.

  In step S304, a prediction method to be applied to the target block is selected from inter-screen prediction and intra-screen prediction. For example, a prediction method that gives a prediction value with a small error for the target block is selected. Alternatively, the encoding process may be actually performed for the two prediction methods, and the smaller evaluation value calculated from the relationship between the generated code amount and the sum of the absolute values of the encoding error images may be selected. The selection information of the selected prediction method is sent to the entropy encoder 111 via the line L112 as information necessary for generating a prediction signal, and is transmitted from the output terminal 112 after being encoded (step S305).

  When the prediction method selected in step S306 is inter-screen prediction, a prediction signal is generated in step S307 based on motion information (motion vector and reference screen information), and the generated inter-screen prediction signal is represented by line L103. And then output to the subtractor 105. In step S308, the motion information is sent to the entropy encoder 111 via the line L112 as information necessary for generating a prediction signal, and is transmitted from the output terminal 112.

  If the prediction method selected in step S306 is intra prediction, a prediction signal is generated in step S309 based on the intra prediction mode, and the generated intra prediction signal is output to the subtractor 105 via the line L103. Is done. In step S310, the intra prediction mode is sent to the entropy encoder 111 via the line L112 as information necessary for generating a prediction signal, and is transmitted from the output terminal 112.

  The encoding method used for the entropy encoder 111 may be arithmetic encoding or variable length encoding.

  FIG. 2 is a block diagram of the video predictive decoding apparatus 200 according to the embodiment of the present invention. As shown in FIG. 2, the moving picture predictive decoding apparatus 200 includes an input terminal 201, a data analyzer 202, an inverse quantizer 203, an inverse transformer 204, an adder 205, a prediction signal generator 208, a frame memory 207, and an output. A terminal 206, a loop filter 209, and a block memory 215 are provided. The inverse quantizer 203 and the inverse transformer 204 correspond to “decoding means” recited in the claims. A decoding means other than those described above may be used. Further, the inverse converter 204 may not be provided. The frame memory 207 corresponds to “image storage means”, and the block memory 215 corresponds to “block storage means”.

  The operation of the video predictive decoding apparatus 200 configured as described above will be described below. The compressed data compressed and encoded by the method described above is input from the input terminal 201. The compressed data includes a residual signal encoded by predicting a target block obtained by dividing an image into a plurality of blocks and information necessary for generating a prediction signal. As shown in FIG. 7, the block size is not limited. A plurality of block sizes and shapes may be mixed on one screen. The decoding order of blocks is described in Non-Patent Document 1, for example. Information necessary for generating the prediction signal includes prediction method selection information and motion information (in the case of inter-screen prediction) or intra prediction mode (in the case of intra-screen prediction).

  The data analyzer 202 decodes the residual signal of the target block, information necessary for generating the prediction signal, and the quantization parameter from the compressed data. The decoded residual signal of the target block is inversely quantized by the inverse quantizer 203 based on the quantization parameter (via the line L202). Further, the inverse quantized residual signal is subjected to inverse discrete cosine transform by the inverse transformer 204, and as a result, the residual signal is restored. Next, information necessary for generating the prediction signal is sent to the prediction signal generator 208 via the line L206. The prediction signal generator 208 generates a prediction signal of the target block based on information necessary for generating the prediction signal. The prediction signal generation processing in the prediction signal generator 208 will be described later with reference to FIG. The generated prediction signal is sent to the adder 205 via the line L208, added to the restored residual signal, regenerates the target block signal, and is output to the loop filter 209 via the line L205. Is stored in the block memory 215 for use in intra prediction. The loop filter unit 209 removes block noise from the reproduction signal input via the line L205, and the reproduction image from which the block noise has been removed is stored in the frame memory 207 as a reproduction image used for decoding and reproduction of the subsequent image. The

  The prediction signal processing flow in the prediction signal generator 208 will be described with reference to FIG. First, in step S402, the prediction method decoded by the data analyzer 202 is acquired.

  When the decoded prediction method is inter-screen prediction (step S403), the motion information (motion vector and reference screen information) decoded by the data analyzer 202 is acquired (step S404), and the frame memory 207 is based on the motion information. To obtain a reference signal from a plurality of reference images and generate a prediction signal (step S405).

  When the decoded prediction method is intra-screen prediction (step S403), the intra prediction mode decoded by the data analyzer 202 is acquired (step S406), the block memory 215 is accessed, and an already reproduced block adjacent to the target block is reproduced. A pixel signal is acquired as a reference sample, and a prediction signal is generated based on the intra prediction mode (step S407). The generated prediction signal is output to the adder 205 via L208.

  The decoding method used for the data analyzer 202 may be arithmetic decoding or variable length decoding.

  Next, the intra prediction method according to the embodiment of the present invention will be described with reference to FIGS. 3 and 7. That is, it is the details of step S309 in FIG. 13 and step S407 in FIG. 14, and using the reference samples acquired from the block memory 113 in FIG. 1 or the block memory 215 in FIG. A method for estimation by extrapolation based on the above will be described.

  In the present invention, in order to suppress the occurrence of noise such as the pseudo contour shown in the problem to be solved by the invention, bilinear interpolation is performed on the reference sample group used for intra prediction for the block that causes the pseudo contour. Apply processing. By slowing the change in the signal of the reference sample group, the appearance of step noise generated at the block boundary of the reference sample group is suppressed.

A bilinear interpolation process applied to the reference sample group will be described with reference to FIG. When the block size of the target block 210 is N × N samples, here, 4N + 1 reference sample groups in the already reproduced signals belonging to the surrounding five already reproduced blocks 220, 230, 240, 250, 260 270 (ref [x] (x = 0 to 4N)) is configured. In the present embodiment, the lower left reference sample BL = ref [0] and the upper right reference sample AR = ref [4N] at the end of the reference sample group 270 and the center of the reference sample group 270, Three upper-left reference samples AL = ref [2N] located on the upper left are defined as key reference samples for bilinear interpolation. At this time, 4N + 1 reference samples are interpolated as follows.
ref '[0] = ref [0] (1)
ref '[i] = BL + (i * (AL-BL) + N) / 2N (i = 1-2N-1) (2)
ref '[2N] = ref [2N] (3)
ref '[2N + i] = AL + (i * (AR-AL) + N) / 2N (i = 1-2N-1) (4)
ref '[4N] = ref [4N] (5)
Here, ref ′ [x] (x = 0 to 4N) indicates the value of the interpolated reference samples after the interpolation process. Expressions (2) and (4) may be modified as Expressions (2) ′ and (4) ′, respectively.
ref '[i] = ((2N-i) * BL + i * AL + N) / 2N (i = 1-2N-1) (2)'
ref '[2N + i] = ((2N-i) * AL + i * AR + N) / 2N (i = 1 to 2N-1) (4)'

  In this way, a reference sample between BL and AL is generated by bilinear interpolation with key reference samples BL and AL, and a reference sample between AL and AR is bilinearly inserted with key reference samples AL and AR. The level of the reference sample value after the interpolation processing adjacent to the target block changes gently. As a result, propagation of block noise to the prediction signal can be suppressed.

  Next, with reference to FIG. 7, a reference sample determination criterion to which bilinear interpolation is applied will be described. In this embodiment, the determination is performed using three key reference samples, two reference samples at the block boundary, and two threshold values. Here, THRESHOLD_ABOVE and THRESHOLD_LEFT are set to the reference sample ref [x] (x = 2N + 1 to 4N-1) at the upper end of the target block and the reference sample ref [x] (x = 1 to 2N-1) at the left end, respectively. The threshold is used for determination for determining whether or not to apply bilinear interpolation. Bilinear interpolation is applied to reference samples that meet the criteria.

In the present embodiment, the following criteria are used. Interpolate_Above and Interpolate_Left in the following two expressions are Boolean values. If the expression on the right side is satisfied, it becomes true (1) and bilinear interpolation is applied. Apply smoothing.
Interpolate_Left = abs (BL + AL-2 * ref [N]) <THRESHOLD_LEFT (6)
Interpolate_Above = abs (AL + AR-2 * ref [3N]) <THRESHOLD_ABOVE (7)
When BL, AL and ref [3N] values are arranged on a straight line, the value of BL + AL−2 * ref [N] is 0. Similarly, when the values of AL, AR, and ref [3N] are arranged on a straight line, the value of AL + AR−2 * ref [3N] is also 0. In other words, the above two formulas show the magnitude of deviation (deviation) of ref [N] for the straight line connecting BL and AL, and the magnitude of deviation (deviation) of ref [3N] for the straight line connecting AL and AR, respectively. It is compared with the threshold value. If the two calculated deviations are smaller than the corresponding threshold THRESHOLD_ABOVE or THRESHOLD_LEFT, the Boolean value (Interpolate_Above or Interpolate_Left) is true, and bilinear interpolation is applied to the reference sample. In equations (6) and (7), abs (x) calculates the absolute value of x.

  At this time, the two threshold values (THRESHOLD_ABOVE and THRESHOLD_LEFT) may be fixed values set in advance, or may be encoded in frame units or slice units in which a plurality of blocks are combined and restored by a decoder. Alternatively, encoding may be performed in block units and restored by a decoder. In FIG. 2, the data analyzer 202 decodes the two threshold values, outputs them to the prediction signal generator 208, and uses them to generate an intra-screen prediction signal described in detail in FIGS. 3 and 4 below.

  FIG. 3 shows a flowchart of processing for estimating an intra prediction sample by extrapolation (direction of intra prediction). First, in step 510, the prediction signal generator (103 or 208, numbers are omitted) from the block memory (113 or 215, numbers are omitted), as shown in the pixel group 270 of FIG. x] (x = 0 to 4N) is acquired. At this time, if the adjacent block has not been reproduced yet for reasons such as the coding order and all of the 4N + 1 reference samples cannot be acquired, the nonexistent sample is padded (the nearby existing sample). Copy the value) and prepare 4N + 1 reference samples. The details of the padding process are described in Non-Patent Document 1. Next, in step 560, two Boolean values Interpolate_Above and Interpolate_Left are calculated based on equations (6) and (7).

  Next, in step 520, the prediction signal generator determines whether the target block satisfies the criteria for applying bilinear interpolation. Specifically, it is determined whether the size of the target block is larger than a predetermined M, and it is determined whether the calculated Interpolate_Above and Interpolate_Left are both true. The reason why the block size is used as a criterion is that a pseudo contour that is a problem is usually generated with a large block size. By setting a large value for M, there is an effect of suppressing unnecessary changes in the reference sample.

If these two criteria are satisfied (block size> = M and Interpolate_Above == true and Interpolate_Left == true), the process proceeds to step 530. Otherwise, the process proceeds to step 540. In step 530, the bilinear interpolation processing shown in equations (1) to (5) is applied to the reference sample ref [x] (x = 0 to 4N), and the interpolated reference samples (interpolated reference samples) ref '[x] (x = 0 ~ 4N) is generated. In step 540, intra smoothing using 121 filters is applied to the reference sample ref [x] (x = 0 to 4N) according to equations (8) and (9).
ref '[i] = ref [i] (i = 0 and 4N) (8)
ref '[i] = (ref [i-1] + 2 * ref [i] + ref [i + 1] +2) / 4 (i = 1 to 4N-1) (9)
Here, ref ′ [x] (x = 0 to 4N) indicates the value of the smoothed reference samples.

  Finally, in step 550, the intra prediction sample of the target block is extrapolated using the already determined intra prediction mode and the reference sample ref ′ [x] (x = 0 to 4N) after interpolation or smoothing. Estimated by the method (direction of prediction in the screen).

  FIG. 4 illustrates FIG. 3 in more detail. Bilinear interpolation and 121 filter switching are performed by changing the left reference sample (ref [x], x = 0 to 2N) and the upper reference sample (ref [x ], x = 2N to 4N), and shows a flowchart of a process of estimating an intra prediction sample by an extrapolation method (direction of intra prediction) in the case of performing independently. First, in step 610, the prediction signal generator (103 or 208, numbers are omitted) from the block memory (113 or 215, numbers are omitted) from the reference sample ref [ x] (x = 0 to 4N) is acquired. At this time, if the adjacent block has not been reproduced yet for reasons such as the coding order and all of the 4N + 1 reference samples cannot be acquired, the nonexistent sample is padded (the nearby existing sample). Copy the value) and prepare 4N + 1 reference samples. The details of the padding process are described in Non-Patent Document 1.

  Next, in step 680, two Boolean values Interpolate_Above and Interpolate_Left are calculated based on equations (6) and (7).

  Next, in step 620, the prediction signal generator determines whether the target block satisfies the criterion for applying bilinear interpolation. Specifically, it is determined whether the size of the target block is larger than a predetermined M, and it is determined whether at least one of the calculated Interpolate_Above and Interpolate_Left is true. If these two criteria are satisfied (block size> = M and Interpolate_Above == true or Interpolate_Left == true), the process proceeds to step 625, and if not, the process proceeds to step 660. In step 660, intra smoothing using a 121 filter is applied to the reference sample group according to equations (8) and (9).

In step 625, it is determined whether the criterion for applying bilinear interpolation of the left reference sample shown in Expression (6) is satisfied. That is, if Interpolate_Left is true (1), the process proceeds to step 630, where the bilinear interpolation processing shown in equations (1) and (2) is applied to the reference sample ref [x] (x = 0 to 2N). Interpolated reference samples (interpolated reference samples) ref ′ [x] (x = 0 to 2N) are generated. If the criterion of Expression (6) is not satisfied, the process proceeds to Step 635, and intra smoothing by 121 filter is performed on the left reference sample ref [x] (x = 0 to 2N) according to Expressions (10) and (11). Apply.
ref '[0] = ref [0] (10)
ref '[i] = (ref [i-1] + 2 * ref [i] + ref [i + 1] +2) / 4 (i = 1 to 2N-1) (11)
Here, ref ′ [x] (x = 0 to 2N) indicates the value of the smoothed reference samples.

Next, in step 640, it is determined whether or not the criterion for applying bilinear interpolation of the upper reference sample shown in Expression (7) is satisfied. In other words, if Interpolate_Above is true (1), the process proceeds to step 650, and the upper reference sample ref [i] (i = 2N + 1 to 4N) is bilinear based on equations (3), (4), and (5). Apply interpolation processing. If the determination criterion of Expression (7) is not satisfied, the process proceeds to Step 655, where the upper reference sample ref [x] (x = 2N + 1 to 4N) is based on Expressions (12), (13), and (14). Apply intra smoothing with 121 filter.
ref '[2N] = ref [2N] (12)
ref '[i] = (ref [i-1] + 2 * ref [i] + ref [i + 1] +2) / 4 (i = 2N + 1 to 4N-1) (13)
ref '[4N] = ref [4N] (14)
Here, ref ′ [x] (x = 2N + 1 to 4N) indicates the value of the smoothed reference samples.

  Finally, in step 670, the intra prediction sample of the target block is excluded using the already determined intra prediction mode and the reference sample ref ′ [x] (x = 0 to 4N) after interpolation or smoothing. Estimate by insertion method (direction of prediction in the screen). For extrapolation, when a line is projected in the direction of intra prediction from the position of the sample in the target block to be extrapolated, to the interpolated or smoothed reference samples, Interpolated or smoothed reference samples after interpolation or smoothing at a position close to the projected line are used.

  A moving picture predictive coding program for causing a computer to function as the moving picture predictive coding apparatus 100 is stored in a recording medium and can be provided. Similarly, a moving picture predictive decoding program for causing a computer to function as the moving picture predictive decoding apparatus 200 can be provided by being stored in a recording medium. Examples of the recording medium include a recording medium such as a USB memory, a flexible disk, a CD-ROM, a DVD, or a ROM, or a semiconductor memory.

  For example, as shown in FIG. 16, the moving picture predictive encoding program P100 includes a block division module P101, a prediction signal generation module P102, a residual signal generation module P103, a residual signal compression module P104, a residual signal restoration module P105, An encoding module P106 and a block storage module P107 are provided.

  For example, as illustrated in FIG. 17, the moving picture predictive decoding program P200 includes a decoding module P201, a prediction signal generation module P202, a residual signal restoration module P203, and a block storage module P204.

  The moving picture predictive encoding program P100 or the moving picture predictive decoding program P200 configured as described above is stored in the recording medium 10 shown in FIGS. 5 and 6 described later, and is executed by a computer described later.

  FIG. 5 is a diagram showing a hardware configuration of the computer 30 for executing the program recorded on the recording medium, and FIG. 6 is an overview diagram of the computer 30 for executing the program stored on the recording medium. is there. The computer 30 here includes a wide range of DVD players, set-top boxes, mobile phones, and the like that have a CPU and perform information processing and control by software.

  As shown in FIG. 6, the computer 30 includes a reading device 12 such as a flexible disk drive device, a CD-ROM drive device, and a DVD drive device, a working memory (RAM) 14 in which an operating system is resident, and a recording medium 10. A memory 16 for storing programs stored therein, a display device 18 such as a display, a mouse 20 and a keyboard 22 as input devices, a communication device 24 for transmitting and receiving data and the like, and a CPU 26 for controlling execution of the programs. And. When the recording medium 10 is inserted into the reading device 12, the computer 30 can access the moving image predictive encoding program stored in the recording medium 10 from the reading device 12, and the moving image predictive encoding program performs the above operation. It becomes possible to operate as the moving image predictive encoding device 100. Similarly, when the recording medium 10 is inserted into the reading device 12, the computer 30 can access the moving image predictive decoding program stored in the recording medium 10 from the reading device 12, and the moving image predictive decoding program performs the above operation. It becomes possible to operate as the moving picture predictive decoding apparatus 200.

  In the present invention, the following modifications are possible.

(A) Criteria for applying bilinear interpolation The criteria for applying bilinear interpolation are not limited to the methods described in the above embodiments. For example, the determination result of interpolation application may always be true, and steps 520, 620, 625, and 640 may be omitted. In this case, an interpolation process is always applied instead of the smoothing process using the 121 filter.

  An intra prediction mode may be added to the determination criterion. For example, since the pseudo contour generated at the block boundary is reduced by the block noise removal process, when the prediction direction of the extrapolation process is vertical or horizontal, the determination result of applying the interpolation process may always be false.

  The block size may be removed from the criterion. Further, instead of the target block block size, the relative relationship between the block size of the target block and the adjacent block may be used as a criterion. In the example of FIG. 7, the block size of the block 260 adjacent to the left of the target block 210 is larger than that of the target block 210. In this case, no block noise occurs around ref [N]. Thus, when the block size of the adjacent block is larger than the target block, the criterion for applying interpolation may be set to false regardless of the result of Expression (6) or (7). On the other hand, adjacent blocks 230, 240, and 250 on the target block 210 are smaller than the target block 210. In this case, since block noise may occur around ref [3N] or ref [2N + N / 2], the interpolation application is determined based on the result of Expression (6) or (7). Note that the relative relationship between the block size of the target block and the adjacent block may be used as a determination criterion together with the block size of the target block.

  The thresholds (THRESHOLD_ABOVE and THRESHOLD_LEFT) in Equations (6) and (7) are individually determined and encoded for different block sizes, block shapes (differences in the vertical and horizontal sizes of blocks) and different intra prediction modes. You may make it restore with. Alternatively, THRESHOLD_ABOVE and THRESHOLD_LEFT may be set to the same value, and only one of them may be encoded and restored by a decoder. In the decoder, the threshold value restored by the data analyzer 202 in FIG. 2 is input to the prediction signal generator 208. The prediction signal generator 208 calculates Interpolate_Above and Interpolate_Left values based on the input threshold (step 560 in FIG. 3 or step 680 in FIG. 4).

  Further, instead of providing determination criteria in steps 520, 620, 625, and 640, the determination result may be included in the bitstream, encoded, and restored by the decoder. In this case, the prediction signal generator 103 of FIG. 1 obtains the Interpolate_Above and Interpolate_Left values (0 or 1) based on the size of the target block and the results of the equations (6) and (7). As the prediction information necessary for prediction, encoding is performed for each block or a block group unit in which a plurality of blocks are collected. That is, the data is sent to the entropy encoder 111 via the line L112, encoded, and sent from the output terminal 112. Note that when obtaining the Interpolate_Above and Interpolate_Left values (0 or 1), the above-described relative relationship between the block size of the target block and the adjacent block, the size of the target block, and the intra prediction mode may be used.

  In the data analyzer 202 shown in FIG. 2, the values of Interpolate_Above and Interpolate_Left are decoded for each block or in units of a block group in which a plurality of blocks are collected, and input to the prediction signal generator 208. The two values may be individually encoded and decoded, or may be encoded and decoded as a set of two values.

  The process of the intra prediction method in the prediction signal generator 208 of FIG. 2 will be described with reference to FIG. In this case, FIG. 15 is replaced with FIG. In FIG. 14, in Step S406, values of Interpolate_Above and Interpolate_Left decoded together with the intra prediction mode are acquired. First, in step 710, the prediction signal generator (103 or 208, the following numbers are omitted) from the block memory (113 or 215, the following numbers are omitted) from a reference sample ref [ x] (x = 0 to 4N) is acquired. At this time, if the adjacent block has not been reproduced yet for reasons such as the coding order and all of the 4N + 1 reference samples cannot be acquired, the nonexistent sample is padded (the nearby existing sample). Copy the value) and prepare 4N + 1 reference samples. The details of the padding process are described in Non-Patent Document 1.

  Next, in Step 790, Interpolate_Above and Interpolate_Left values are acquired. In step 720, the prediction signal generator determines whether one of Interpolate_Above and Interpolate_Left is 1. If any value is 1, the process proceeds to step 725, and if not, the process proceeds to step 760. In step 760, intra smoothing using a 121 filter is applied to the reference sample group according to equations (8) and (9).

  In step 725, if the value of Interpolate_Left is 1, the process proceeds to step 730, where the bilinear interpolation processing shown in equations (1) and (2) is applied to the reference sample ref [x] (x = 0 to 2N), Interpolated reference samples (interpolated reference samples) ref '[x] (x = 0 to 2N) are generated. If the value of Interpolate_Left is 0, the process proceeds to step 735, and intra smoothing using 121 filters is applied to the left reference sample ref [x] (x = 0 to 2N) according to equations (10) and (11).

  Next, in Step 740, if the Interpolate_Above value is 1, the process proceeds to Step 750, and the upper reference sample ref [i] (i = 2N + 1 to 4N) is calculated based on Expressions (3), (4), and (5). Apply bilinear interpolation to. If the Interpolate_Above value is 0, the process proceeds to step 755, and intra smoothing using the 121 filter is performed on the left reference sample ref [x] (x = 2N + 1 to 4N) based on the equations (12), (13), and (14). Apply.

  Finally, in step 770, the intra prediction sample of the target block is extrapolated using the decoded intra prediction mode and the reference sample ref ′ [x] (x = 0 to 4N) after interpolation or smoothing. Estimated by the method (direction of prediction in the screen).

(B) Interpolation process In the above description, bilinear interpolation is used for the interpolation process. However, any other interpolation process may be used as long as noise at the block boundary can be removed. For example, all reference samples may be replaced with an average value of key reference samples. The interpolation processing method may be switched depending on the block size or the intra prediction type, and the applied interpolation processing method may be included in the bitstream for encoding and decoding.

(C) Process flow of intra prediction of reference sample The process flow for estimating an intra prediction sample by extrapolation (direction of intra prediction) is not limited to the procedure in FIG. For example, steps 625, 630 and 635 may be reversed in order with steps 640, 650 and 655. Moreover, you may implement Formula (3) and Formula (12) by step 630,635 instead of step 650,655. In addition, since the processing results of the expressions (1), (3), and (5) are the same as those of the expressions (10), (12), and (14), either immediately before step 625 (between steps 620 and 625) or between steps 650 and 655 Immediately after that (between step 650 or 655 and step 670) may be performed together.

Further, the determination criterion in step 620 may be only the block size. At this time, if Expression (12) is replaced with Expressions (15) and (16), the processing result will be the same as in FIG.
ref '[2N] = ref [2N]
if Interpolate_Above == true || Interpolate_Left == true (15)
ref '[2N] = (ref [2N-1] + 2 * ref [2N] + ref [2N + 1] +2) / 4 others (16)
Here, ref ′ [2N] indicates the value of the smoothed reference samples.

(D) Block size In the above description, the target block is a square block. However, the interpolation processing to the reference sample of the present invention can be similarly applied to a non-square block. An example in which the target block 290 block size is N × 2N is shown in FIG. In this case, the number of ref [x] is 3N + 1.

(E) Key Reference Sample In the above description, the key reference samples are the three at the end and the center of the reference sample group, but the number and position are not limited. For example, the number and position may be changed according to the size of the reference block and the relative relationship between the reference block and the adjacent block, and the number and position of the key reference samples are included in the bitstream for encoding and decoding. Also good. Alternatively, the key reference samples may be encoded and decoded as instruction information indicating whether the default or the other key reference sample is to be used, with the default three at the end and the center of the reference sample group. In the data analyzer 202 of FIG. 2, the key reference sample is updated. As a key reference sample to be updated, ref [N + N / 2] and ref [2N + N / 2] in FIG. 7 may be added, or they may be used instead of ref [2N]. Also, instead of ref [0] and ref [4N], ref [N / 2] and ref [3N + N / 2] are used, and ref [1] to ref [N / 2-1] and ref [3N + 121 filters may be applied to N / 2] to ref [4N-1].

(F) Formula of determination criteria The determination formulas used in steps 520, 620, 625, and 640 are not limited to formulas (6) and (7). For example, ref [N + 1] or ref [3N + 1] may be used instead of ref [N] and ref [3N] in FIG.

  A video predictive coding apparatus according to an aspect of the present invention includes: a block dividing unit that divides an input image into a plurality of blocks; and a target block that is an encoding target among blocks divided by the block dividing unit. Prediction signal generating means for generating an intra-screen prediction signal of a highly correlated block using a reference sample already reproduced adjacent to the target block, and a residual between the prediction signal of the target block and the pixel signal of the target block A residual signal generating means for generating a signal, a residual signal compressing means for compressing the residual signal generated by the residual signal generating means, and a reproduction residual signal in which the compressed data of the residual signal is restored A residual signal restoring means, an encoding means for encoding compressed data of the residual signal, and adding the prediction signal and the reproduction residual signal to add the target block. And a block storage unit that stores the restored pixel signal of the target block to be used as the reference sample, and the prediction signal generation unit is stored in the block storage unit. A reference sample is acquired from the already-reproduced blocks around the target block, and an interpolation process is performed between two or more key reference samples at predetermined positions in the reference sample to generate an interpolation reference sample. Determining an intra prediction mode, generating the intra prediction by extrapolating the interpolation reference sample based on the determined intra prediction mode, and encoding means including the intra prediction mode in compressed data It is characterized by becoming.

  In the above moving picture predictive encoding device, the prediction signal generation means applies the reference sample interpolation process and the reference sample smoothing process based on a comparison between the key reference sample and a predetermined threshold value. You may switch to and carry out.

  In the above moving picture predictive coding apparatus, the key reference sample is set to three reference samples positioned at the end and the center of the reference sample group, and the interpolation processing is performed in a bilinear order with respect to the reference samples between the key reference samples. It may be an insertion process.

  A video predictive decoding apparatus according to an aspect of the present invention includes an intra prediction mode and a residual that indicate an intra-screen prediction method for a target block to be decoded from compressed data that is encoded by being divided into a plurality of blocks. Decoding means for decoding the compressed data of the signal, prediction signal generating means for generating an intra-screen prediction signal using the intra-prediction mode and the already reproduced reference sample adjacent to the target block, and the residual signal Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data, and the pixel signal of the target block is restored by adding the prediction signal and the reproduction residual signal, and the restored Block storage means for storing the pixel signal of the target block for use as the reference sample, and the prediction signal generation means includes the block storage means. In order to obtain reference samples from the already-reproduced blocks around the target block stored in and to generate an interpolated reference sample, two or more key reference samples at predetermined positions are generated in the reference sample. An intra-screen prediction signal is generated by extrapolating the interpolation reference sample based on the intra prediction mode.

  In the above moving picture predictive decoding apparatus, the prediction signal generation means applies the reference sample interpolation process and the reference sample smoothing process based on a comparison between the key reference sample and a predetermined threshold. You may switch and implement.

  In the above moving picture predictive decoding apparatus, the key reference sample is set to three reference samples located at the end and the center of the reference sample group, and the interpolation process performs bilinear interpolation on the reference samples between the key reference samples. It may be a process.

  The present invention can also be regarded as an invention related to a video predictive encoding method, an invention related to a video predictive decoding method, an invention related to a video predictive encoding program, an invention related to a video predictive decoding program, and is as follows. Can be described in

  A video predictive encoding method according to an aspect of the present invention is a video predictive encoding method executed by a video predictive encoding device, the block dividing step of dividing an input image into a plurality of blocks, Prediction signal generation for generating an intra-screen prediction signal of a block having a high correlation with the target block to be encoded among the blocks divided by the block division step, using already reproduced reference samples adjacent to the target block A residual signal generating step for generating a residual signal between the prediction signal of the target block and a pixel signal of the target block, and a residual signal for compressing the residual signal generated by the residual signal generating step A compression step, a residual signal restoration step for generating a reproduction residual signal obtained by restoring the compressed data of the residual signal, and compressed data of the residual signal In order to restore the pixel signal of the target block by adding the encoding step of encoding, the prediction signal, and the reproduction residual signal, and use the restored pixel signal of the target block as the reference sample A block storing step for storing the reference sample, wherein in the prediction signal generating step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference is generated to generate an interpolated reference sample. Interpolating between two or more key reference samples at predetermined positions among samples, determining an intra prediction mode, extrapolating the interpolation reference sample based on the determined intra prediction mode, and performing intra prediction A signal is generated, and in the encoding step, the intra prediction mode is included in the compressed data and encoded. That.

  A moving image predictive decoding method according to an aspect of the present invention is a moving image predictive decoding method executed by a moving image predictive decoding device, and decodes from compressed data divided into a plurality of blocks and encoded. A decoding step for decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be processed and compressed data of a residual signal, and the intra-prediction mode and a replayed reference sample adjacent to the target block A prediction signal generation step for generating an intra-screen prediction signal, a residual signal restoration step for restoring the reproduction residual signal of the target block from the compressed data of the residual signal, and the prediction signal and the reproduction residual signal. The pixel signal of the target block is restored by adding, and the restored pixel signal of the target block is stored for use as the reference sample. A block storage step, wherein in the prediction signal generation step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference sample is generated to generate an interpolation reference sample. An interpolation process is performed between two or more key reference samples at predetermined positions, and the intra prediction signal is generated by extrapolating the interpolation reference samples based on the intra prediction mode.

  A moving picture predictive encoding program according to an aspect of the present invention includes a block dividing unit that divides an input image into a plurality of blocks, and a target that is an encoding target among blocks divided by the block dividing unit. A prediction signal generating means for generating an intra-screen prediction signal of a block having a high correlation with the block by using a previously reproduced reference sample adjacent to the target block; a prediction signal of the target block; and a pixel signal of the target block; A residual signal generating means for generating the residual signal, a residual signal compressing means for compressing the residual signal generated by the residual signal generating means, and a reproduction residual obtained by restoring the compressed data of the residual signal A residual signal restoring means for generating a signal, an encoding means for encoding compressed data of the residual signal, and adding the prediction signal and the reproduced residual signal Therefore, a motion picture predictive encoding program for causing the pixel signal of the target block to be restored and functioning as block storage means for storing the restored pixel signal of the target block for use as the reference sample, The prediction signal generation means obtains reference samples from the already reproduced blocks around the target block stored in the block storage means, and generates a reference position of the reference samples in order to generate an interpolation reference sample. Interpolating between two or more key reference samples, determining an intra prediction mode, extrapolating the interpolation reference samples based on the determined intra prediction mode, generating the intra prediction signal, and encoding The means is characterized by encoding the intra prediction mode by including it in the compressed data.

  A moving picture predictive decoding program according to an aspect of the present invention is an intra prediction mode showing an intra-screen prediction method of a target block to be decoded from compressed data encoded by dividing a computer into a plurality of blocks. Decoding means for decoding the compressed signal of the residual signal, prediction signal generating means for generating an intra-screen prediction signal using the intra prediction mode and the already-reproduced reference samples adjacent to the target block, and the residual signal Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data of the difference signal, and the pixel signal of the target block is restored and restored by adding the prediction signal and the reproduction residual signal Moving picture prediction for functioning as block storage means for storing the pixel signal of the target block that has been used for use as the reference sample The prediction signal generation means obtains a reference sample from the already-reproduced block around the target block stored in the block storage means, and generates the interpolated reference sample to generate the interpolation reference sample. Of these, two or more key reference samples at predetermined positions are interpolated, and the interpolated reference samples are extrapolated based on the intra prediction mode to generate the intra prediction signal.

  A video predictive coding apparatus according to an aspect of the present invention includes: a block dividing unit that divides an input image into a plurality of blocks; and a target block that is an encoding target among blocks divided by the block dividing unit. Prediction signal generating means for generating an intra-screen prediction signal of a highly correlated block using a reference sample already reproduced adjacent to the target block, and a residual between the prediction signal of the target block and the pixel signal of the target block A residual signal generating means for generating a signal, a residual signal compressing means for compressing the residual signal generated by the residual signal generating means, and a reproduction residual signal in which the compressed data of the residual signal is restored A residual signal restoring means, an encoding means for encoding compressed data of the residual signal, and adding the prediction signal and the reproduction residual signal to add the target block. And a block storage unit that stores the restored pixel signal of the target block to be used as the reference sample, and the prediction signal generation unit is stored in the block storage unit. A reference sample is acquired from the already-reproduced blocks around the target block, and an interpolation process is performed between two or more key reference samples at predetermined positions in the reference sample to generate an interpolation reference sample. Determining an intra prediction mode, generating the intra prediction signal by extrapolating the interpolated reference sample based on the determined intra prediction mode, and the encoding means including the intra prediction mode in compressed data And the prediction signal generating means generates the reference sample based on a comparison between a value based on the key reference sample and a predetermined threshold. Which comprises carrying out switching between the smoothing processing of the interpolation process to the reference sample application basis.

  A video predictive decoding apparatus according to an aspect of the present invention includes an intra prediction mode and a residual that indicate an intra-screen prediction method for a target block to be decoded from compressed data that is encoded by being divided into a plurality of blocks. Decoding means for decoding the compressed data of the signal, prediction signal generating means for generating an intra-screen prediction signal using the intra-prediction mode and the already reproduced reference sample adjacent to the target block, and the residual signal Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data, and the pixel signal of the target block is restored by adding the prediction signal and the reproduction residual signal, and the restored Block storage means for storing the pixel signal of the target block for use as the reference sample, and the prediction signal generation means includes the block storage means. In order to obtain reference samples from the already-reproduced blocks around the target block stored in and to generate an interpolated reference sample, two or more key reference samples at predetermined positions are generated in the reference sample. Interpolating and extrapolating the interpolation reference sample based on the intra prediction mode to generate the intra prediction signal, and the prediction signal generating means includes a value based on the key reference sample and a predetermined threshold value. On the basis of the comparison, the reference sample interpolation process and the reference sample smoothing process are switched by being applied.

  A video predictive encoding method according to an aspect of the present invention is a video predictive encoding method executed by a video predictive encoding device, the block dividing step of dividing an input image into a plurality of blocks, Prediction signal generation for generating an intra-screen prediction signal of a block having a high correlation with the target block to be encoded among the blocks divided by the block division step, using already reproduced reference samples adjacent to the target block A residual signal generating step for generating a residual signal between the prediction signal of the target block and a pixel signal of the target block, and a residual signal for compressing the residual signal generated by the residual signal generating step A compression step, a residual signal restoration step for generating a reproduction residual signal obtained by restoring the compressed data of the residual signal, and compressed data of the residual signal In order to restore the pixel signal of the target block by adding the encoding step of encoding, the prediction signal, and the reproduction residual signal, and use the restored pixel signal of the target block as the reference sample A block storing step for storing the reference sample, wherein in the prediction signal generating step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference is generated to generate an interpolated reference sample. Interpolating between two or more key reference samples at predetermined positions among samples, determining an intra prediction mode, extrapolating the interpolation reference sample based on the determined intra prediction mode, and performing intra prediction A signal is generated, and in the encoding step, the intra prediction mode is encoded by including the intra prediction mode in compressed data, and the prediction signal is generated. In the forming step, based on a comparison between a value based on the key reference sample and a predetermined threshold, the reference sample interpolation process and the reference sample smoothing process are switched by being applied. To do.

A moving image predictive decoding method according to an aspect of the present invention is a moving image predictive decoding method executed by a moving image predictive decoding device, and decodes from compressed data divided into a plurality of blocks and encoded. A decoding step for decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be processed and compressed data of a residual signal, and the intra-prediction mode and a replayed reference sample adjacent to the target block A prediction signal generation step for generating an intra-screen prediction signal, a residual signal restoration step for restoring the reproduction residual signal of the target block from the compressed data of the residual signal, and the prediction signal and the reproduction residual signal. The pixel signal of the target block is restored by adding, and the restored pixel signal of the target block is stored for use as the reference sample. A block storage step, wherein in the prediction signal generation step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference sample is generated to generate an interpolation reference sample. Interpolating between two or more key reference samples at predetermined positions, generating the intra prediction signal by extrapolating the interpolation reference sample based on the intra prediction mode, and in the prediction signal generating step The reference sample interpolation process and the reference sample smoothing process are applied and switched based on a comparison between a value based on the key reference sample and a predetermined threshold value.

  A video predictive decoding apparatus according to an aspect of the present invention includes an intra prediction mode and a residual that indicate an intra-screen prediction method for a target block to be decoded from compressed data that is encoded by being divided into a plurality of blocks. Decoding means for decoding the compressed data of the signal, prediction signal generating means for generating an intra-screen prediction signal using the intra-prediction mode and the already reproduced reference sample adjacent to the target block, and the residual signal Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data, and the pixel signal of the target block is restored by adding the prediction signal and the reproduction residual signal, and the restored Block storage means for storing the pixel signal of the target block for use as the reference sample, and the prediction signal generation means includes the block storage means. In order to obtain reference samples from the already-reproduced blocks around the target block stored in and to generate an interpolated reference sample, two or more key reference samples at predetermined positions are generated in the reference sample. Interpolating, extrapolating the interpolated reference sample based on the intra prediction mode to generate the intra prediction signal, the key reference sample as three reference samples located at the end and center of the reference sample group, The prediction signal generation means is configured to calculate a difference between an added value of two key reference samples located at one end and the center of the reference sample group and a double value of a reference sample located at the center of the two key reference samples. Of the absolute value of the reference sample group and a predetermined threshold value, the addition value of the two key reference samples located at the other end and the center of the reference sample group, and the two keys. Based on the result of comparing the absolute value of the difference with the value twice the reference sample located in the center of the reference sample, the threshold value, and the size of the target block, the interpolation processing of the reference sample, It is characterized in that the reference sample smoothing process is switched by application.

  A moving image predictive decoding method according to an aspect of the present invention is a moving image predictive decoding method executed by a moving image predictive decoding device, and decodes from compressed data divided into a plurality of blocks and encoded. A decoding step for decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be processed and compressed data of a residual signal, and the intra-prediction mode and a replayed reference sample adjacent to the target block A prediction signal generation step for generating an intra-screen prediction signal, a residual signal restoration step for restoring the reproduction residual signal of the target block from the compressed data of the residual signal, and the prediction signal and the reproduction residual signal. The pixel signal of the target block is restored by adding, and the restored pixel signal of the target block is stored for use as the reference sample. A block storage step, wherein in the prediction signal generation step, a reference sample is obtained from a previously reconstructed block around the stored target block, and the reference sample is generated to generate an interpolation reference sample. Interpolating between two or more key reference samples at predetermined positions, generating the intra prediction signal by extrapolating the interpolation reference samples based on the intra prediction mode, refer to the key reference samples Three reference samples located at the end and center of the sample group are used, and in the prediction signal generation step, the sum of two key reference samples located at one end and the center of the reference sample group and the two key reference samples A result of comparing an absolute value of a difference with a value twice as large as that of the reference sample located in the center of the reference sample and a predetermined threshold value, and the reference sample group The result of comparing the threshold value with the absolute value of the difference between the sum of the two key reference samples located at the other end and the center and the double value of the reference sample located at the center of the two key reference samples And, based on the size of the target block, the reference sample interpolation processing and the reference sample smoothing processing are applied and switched appropriately.

  DESCRIPTION OF SYMBOLS 100 ... Moving image predictive coding apparatus, 101 ... Input terminal, 102 ... Block divider, 103 ... Prediction signal generator, 104 ... Frame memory, 105 ... Subtractor, 106 ... Converter, 107 ... Quantizer, 108 ... Inverse quantizer, 109 ... Inverse transformer, 110 ... Adder, 111 ... Entropy encoder, 112 ... Output terminal, 113 ... Block memory, 114 ... Loop filter, 200 ... Moving picture predictive decoding apparatus, 201 ... Input Terminals 202, data analyzer 203, inverse quantizer 204, inverse transformer, 205, adder, 206, output terminal, 207, frame memory, 208, prediction signal generator, 209, loop filter, 215 ... block memory.

Claims (2)

Decoding means for decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be decoded and compressed data of a residual signal from compressed data divided into a plurality of blocks and encoded,
Prediction signal generating means for generating an intra-screen prediction signal using the intra prediction mode and the already-reproduced reference sample adjacent to the target block;
Residual signal restoration means for restoring the reproduction residual signal of the target block from the compressed data of the residual signal;
Block storage means for restoring the pixel signal of the target block by adding the prediction signal and the reproduction residual signal, and storing the restored pixel signal of the target block for use as the reference sample; Comprising
The prediction signal generating means includes
A reference sample is acquired from the already played blocks around the target block stored in the block storage means,
Interpolating between two or more key reference samples at a predetermined position among the reference samples to generate an interpolation reference sample;
Extrapolating the interpolated reference sample based on the intra prediction mode to generate the intra prediction signal,
The key reference samples are three reference samples located at the end and center of the reference sample group,
The prediction signal generating means includes
The first absolute value of the difference between the sum of the two key reference samples located at one end and the center of the reference sample group and the double value of the reference sample located at the center of the two key reference samples A result of comparing the absolute value with a predetermined threshold;
The second absolute value of the difference between the sum of the two key reference samples located at the other end and the center of the reference sample group and the double value of the reference sample located at the center of the two key reference samples A result of comparing the absolute value with the threshold;
The size of the target block;
Based on the above, the reference sample interpolation processing and the reference sample smoothing processing are applied and switched,
The prediction signal generation means performs the interpolation processing of the reference sample when the first absolute value is smaller than the threshold value and the second absolute value is smaller than the threshold value,
The threshold value is restored by the moving picture predictive decoding apparatus,
A video predictive decoding apparatus characterized by the above. A video predictive decoding method executed by a video predictive decoding device,
A decoding step of decoding an intra prediction mode indicating an intra-screen prediction method of a target block to be decoded and compressed data of a residual signal from among the compressed data encoded by being divided into a plurality of blocks;
A prediction signal generation step of generating an intra-screen prediction signal using the intra prediction mode and the already reproduced reference sample adjacent to the target block;
A residual signal restoration step of restoring the reproduction residual signal of the target block from the compressed data of the residual signal;
A block storing step of restoring the pixel signal of the target block by adding the prediction signal and the reproduction residual signal, and storing the restored pixel signal of the target block for use as the reference sample; Comprising
In the predicted signal generation step,
A reference sample is obtained from the already played blocks around the stored target block,
Interpolating between two or more key reference samples at a predetermined position among the reference samples to generate an interpolation reference sample;
Extrapolating the interpolated reference sample based on the intra prediction mode to generate the intra prediction signal,
The key reference samples are three reference samples located at the end and center of the reference sample group,
In the predicted signal generation step,
The first absolute value of the difference between the sum of the two key reference samples located at one end and the center of the reference sample group and the double value of the reference sample located at the center of the two key reference samples A result of comparing the absolute value with a predetermined threshold;
The second absolute value of the difference between the sum of the two key reference samples located at the other end and the center of the reference sample group and the double value of the reference sample located at the center of the two key reference samples A result of comparing the absolute value with the threshold;
The size of the target block;
Based on the above, the reference sample interpolation processing and the reference sample smoothing processing are applied and switched,
In the predicted signal generation step, when the first absolute value is smaller than the threshold value and the second absolute value is smaller than the threshold value, an interpolation process of the reference sample is performed,
The threshold value is restored by the moving picture predictive decoding apparatus,
A video predictive decoding method characterized by the above.

Images