JP4184249B2 - Image encoding device - Google Patents

Description

  The present invention relates to an image encoding device, and more particularly to an image encoding device that encodes still image and moving image data.

  As a next-generation moving image encoding method, a technique described in Non-Patent Document 1 below, which is a moving image encoding method aiming at further improvement in encoding efficiency compared to the conventional method, is known. In the present technology, encoding efficiency is improved by an image encoding technique called intra prediction as compared with a conventional moving image encoding system.

  Hereinafter, intra prediction used in the present technology will be described. In the present technology, an image is divided into a plurality of blocks and encoding is performed for each block. Intra prediction is a prediction method for creating a predicted image of an image of a block on the basis of an image of a neighboring block that has already been encoded with respect to an image of a block to be encoded. Encoding efficiency is improved by encoding a difference image between an appropriate prediction image created by intra prediction and an image to be encoded.

  Further, in the present technology, as illustrated in FIG. 10A, the image to be encoded is divided into a plurality of blocks each having a size of 16 × 16 pixels. Each block is further divided into 16 equal parts as shown in FIG. 10B and divided into sub-blocks of 4 × 4 pixel units. In the following description, each pixel value in the block is represented as Xijk. Here, i and j are horizontal and vertical positions (0 ≦ i, j <4) in the sub-block, respectively, and k is a sub-block identification ID (0 indicating each sub-block in the block shown in FIG. 10B). ≦ k <16).

  FIG. 11 shows a case where a sub-block of 4 × 4 pixels is used as a unit as an example of a prediction method in intra prediction used in the present technology.

  The prediction method used in intra prediction is not one type, but in the case of a 4 × 4 pixel block, there are nine prediction methods a to i as shown in FIG. The arrows in FIG. 11A schematically represent the difference in the prediction direction of each prediction method, and this will be described using the examples of FIGS. 11B to 11E.

  FIGS. 11B to 11E illustrate the relationship between the predicted images of the sub-blocks created by the prediction methods a to d, and the relationship of the nine pixels belonging to the peripheral blocks that are referred to when the predicted images are created. FIG.

  In addition, the numerical value in a figure represents the pixel value of the said pixel. For example, in the case of the prediction method a, as shown by the arrow in the horizontal direction in FIG. 11A, in FIG. 11B, prediction in the horizontal direction is performed based on the four pixels adjacent to the left side of the sub-block. A prediction image has been created.

  Similarly, in the prediction method b, as shown in FIGS. 11A and 11C, a prediction image is created based on four pixels adjacent on the upper side by vertical prediction, and in the prediction method d, FIG. As shown in a) and FIG. 11 (e), a prediction image is created based on the adjacent nine pixels by the oblique prediction.

  However, unlike the other prediction methods, the prediction method c is an exceptional prediction method that does not have a specific prediction direction. As shown in FIG. 11D, a prediction image is created by the average value of adjacent nine pixels. . The prediction methods other than the prediction methods a to d shown in FIGS. 11B to 11E are not described because a prediction image according to the prediction direction is created by the same method.

  Note that there are nine prediction methods prepared from a to i for the luminance component 4 × 4 pixel sub-block, four for the luminance component 16 × 16 pixel block, and four for the color difference component block. There are many combinations of block size and prediction direction as methods.

  In the conventional image coding apparatus, as shown in the operation flow of FIG. 12, one of a plurality of prediction methods is selected, the generated code amount in the prediction method is estimated (S1), and all predictions are similarly performed. The estimation of the generated code amount is repeated for the method (S2), the prediction method that is expected to have the smallest generated code amount is selected from the estimation result of the generated code amount, and encoding is performed (S3). -S3 is repeated (S4).

  For the estimation of the amount of generated code, a method of actually encoding by the prediction method and measuring the generated code amount or a method of estimating from the magnitude of error between the encoding target image and the predicted image is used. .

  When estimation is performed using the error ek between the encoding target image and the predicted image, the error ek is expressed by the following equation, where Xijk is the pixel value of the image of the sub-block and Yijk is the corresponding pixel value of the predicted image. Is done.

In the above estimation, the sum of absolute differences (n = 1) or the sum of squared differences (n = 2) is usually used. Note that the generated code amount is estimated as the generated code amount is small as the error e is small and the generated code amount is large as the error e is large.
Final Committee Draft ISO / IEC 14496-10

  In the conventional image encoding apparatus, since an optimal prediction method is selected from among a plurality of prediction methods, it is necessary to estimate the amount of generated codes and select an optimal one for all prediction methods, which is necessary for encoding. There was a problem that the amount of calculation became enormous.

  The present invention has been made to solve the above-described problems, and an object thereof is to provide an image encoding apparatus that can reduce the amount of calculation required for encoding.

  The present invention has been made in order to achieve the above object, and the first image encoding device of the present invention provides a block-divided image for each block or each sub-block obtained by further dividing the block. An image encoding apparatus that performs intraframe prediction encoding using a plurality of changeable prediction methods, wherein a part of the plurality of prediction methods is set as a prediction method candidate under a predetermined condition, and the block from the prediction method candidate It has a feature that determines the prediction method.

Further , the predetermined condition is an image compression rate, and when the compression rate is high, the ratio of the prediction method for each block in the prediction method candidate is high, and when the compression rate is low, A feature of increasing the ratio of the prediction method for each sub-block in the prediction method candidates is provided.

  The first image encoding apparatus of the present invention selects a prediction method candidate based on a predetermined condition, and therefore does not require code amount estimation in all prediction methods for determining a prediction method, and for determining a prediction method. It is possible to reduce the amount of computation of

It is also possible to select an appropriate prediction method candidate using the correlation between the prediction method and the image compression rate.

  An image encoding device according to an example of an embodiment of the present invention will be described in detail below.

  An image encoding apparatus according to an example of an embodiment of the present invention divides an input image into a plurality of blocks and performs encoding for each block. Furthermore, the image encoding apparatus selects one prediction method from a plurality of prediction methods determined in advance for each of the divided blocks, and creates a prediction image for the block from images belonging to adjacent blocks based on the selected prediction method. . Alternatively, the block is further divided into a plurality of sub-blocks, one prediction method is selected from a plurality of prediction methods predetermined for each sub-block, and the sub-block is selected from images belonging to adjacent sub-blocks based on the selected prediction method. Create a predicted image for.

  Further, the image encoding device generates a difference image between the input image of the block or the sub-block and the prediction image, and performs encoding of the difference image and the selected prediction method type. As such an encoding method, there is the technique of Non-Patent Document 1 described above.

  In this technique, four types of prediction methods in units of 16 × 16 pixels are used, and nine types of prediction methods in units of 4 × 4 pixels are used. The encoding method of the present invention uses these blocks, subblock sizes, and prediction methods. However, the present invention is not limited to this, and any encoding scheme can be applied as long as it is as described above. In the following embodiment, the case of the block and sub block shown in FIG. 10 will be described.

  FIG. 1 is a block diagram showing a configuration of an image encoding device according to the first embodiment of the present invention.

  In FIG. 1, 11 is a prediction method control unit that determines a prediction method for determining an appropriate prediction method from a plurality of prediction methods, and 12 is a prediction image that generates a prediction image based on the prediction method determined by the prediction method control unit 11. A generating unit, 13 is a difference image generating unit that generates a difference image by inputting a prediction image and an image to be encoded, 14 is an image encoding unit that encodes the type of the difference image and the prediction method, and 15 is a difference image. An image synthesis unit for synthesizing the predicted image, 16 an image accumulation unit for accumulating moving images, 17 a recording mode selection unit for determining a recording mode of the image coding apparatus by a user operation, and 18 for recording and accumulating encoded data. It is a recording medium such as a magnetic disk, a removable magneto-optical disk, or a semiconductor memory card.

  Note that a configuration may be employed in which communication means is provided instead of the recording medium 18 and the encoded data is transmitted through the communication means.

  An encoding operation by the image encoding apparatus according to the present embodiment will be described with reference to the operation flow of FIG.

  The image encoding apparatus includes a plurality of recording modes in which the recording image size, recording quality, and the like are parameterized, and the user determines the recording mode of the image using the recording mode selection unit 17 in advance. To do.

  When encoding is started, first, the prediction method control unit 11 determines a prediction method candidate used for the image (S1).

  The prediction method candidate is determined by comparing control information based on the recording mode of the image determined in advance by the user with a predetermined threshold value.

  For example, a case where the resolution of the image is used as control information will be described. However, in the following description, the number of pixels constituting the image is used as an encoding parameter expressing the resolution, and the larger the number of pixels In, the higher the resolution. The prediction method candidate is determined as follows using the number of pixels In and a predetermined threshold value TH1.

  When In> TH1, all prediction methods in units of blocks are set as prediction method candidates for the image.

  In the case of In ≦ TH1, all prediction methods in units of sub-blocks are set as prediction method candidates for the image.

  As In, the number of horizontal pixels of luminance data in the image data or the total number of pixels (the number of horizontal pixels × the number of vertical pixels) can be used.

  In general, an image with a low resolution has a lower correlation between adjacent pixels than an image with a high resolution, so that prediction efficiency in a block-based prediction method is lowered. On the other hand, when the resolution is high, the difference in prediction efficiency between the prediction methods for each block and each sub-block is small. Therefore, the coding efficiency is better when the block-based prediction method is used because the amount of code indicating the prediction method can be suppressed. Therefore, it is possible to select an appropriate prediction method candidate according to the image resolution by previously determining an appropriate threshold value TH1 through experiments.

Also, instead of using only one block unit or sub-block unit as a prediction method candidate, a plurality of threshold values are used to determine the ratio of the block-unit and sub-block unit prediction methods to the prediction method candidate. A method of switching in stages may be used. For example, using threshold values TH1 and TH2,
When In> TH1, all prediction methods in units of blocks are set as prediction method candidates for the image.

  In the case of TH2 <In ≦ TH1, in general, a prediction method with high prediction efficiency is selected by half each of all prediction methods in block units and sub-block units, respectively, and set as a prediction method candidate for the image.

  By switching the prediction method candidates in three stages, it is possible to select a more appropriate prediction method candidate according to the image resolution.

Further, prediction method candidates may be determined based on control information related to recording modes such as a high-quality recording mode and a long-time recording mode. That is, it is possible to determine a prediction method candidate using the allocated code amount Ic per block as an encoding parameter corresponding to the recording mode and using the predetermined threshold value TH3. For example,
If Ic <TH3, all prediction methods in block units are set as prediction method candidates for the image.

  If Ic ≧ TH3, all prediction methods in units of sub-blocks are set as prediction method candidates for the image.

  When the allocated code amount per block is small, the quality of an image generated by encoding is reduced, so that the prediction accuracy is reduced and the difference in prediction efficiency between the prediction methods is reduced. In addition, the ratio of the code amount indicating the prediction method in the entire code amount is relatively high. Therefore, when the allocated code amount per block is small, by selecting a prediction method for each block, it is possible to suppress the code amount necessary for encoding the prediction method and increase the encoding efficiency. On the other hand, when the allocated code amount per block is large, it is possible to increase the encoding efficiency by selecting a prediction method in units of sub-blocks with relatively high prediction accuracy.

  Therefore, it is possible to select an appropriate prediction method candidate corresponding to the allocated code amount per block by previously determining an appropriate threshold value TH3 through experiments. As in the case of using the image resolution, it is also possible to select the prediction method candidates step by step using a plurality of threshold values.

In the above description, the allocated code amount Ic per block is used as the encoding parameter corresponding to the recording mode. However, the compression rate Ir that is the inverse thereof may be used. In that case, the prediction method candidate determination method is, for example,
If Ir> TH3 ′, all prediction methods in block units are set as prediction method candidates for the image.

  If Ir ≦ TH3 ′, all prediction methods for each sub-block are set as prediction method candidates for the image.

  It is realizable by the structure of.

  Next, a prediction method to be used for the block to be encoded is determined from the selected prediction method candidates (S2). In addition, when using the prediction method per subblock, the prediction method used for every subblock is determined. Any method may be used as a method for determining a prediction method used in the block from the selected prediction method candidates. For example, a prediction method that estimates the generated code amount for each prediction method candidate and minimizes the generated code amount can be used.

  Note that this step can be omitted when only one type of prediction method is selected as a prediction method candidate.

  Next, the block is encoded based on the determined prediction method (S3). The predicted image generation unit 11 refers to the encoded image stored in the image storage unit 16 based on the designated prediction method, and generates a predicted image of the block or the subblock. The difference image generation unit 13 creates a difference image between the image to be encoded of the block or the subblock and the prediction image, and the image encoding unit 14 determines the type of the prediction method and the difference image of the block or the subblock. Encoding is performed, and the encoded data is sequentially stored in the recording medium 18. The encoded difference image is synthesized with the predicted image by the image synthesizing unit 15 and accumulated in the image accumulating unit 16 as an encoded image of the block or the sub-block, and used for prediction of other blocks and sub-blocks. It is done.

  The operations of S2 to S3 are repeated for all the blocks of the encoding target image (S4), and the encoding of the image is completed.

  In this way, the encoded data stored in the recording medium 18 according to the present invention has the following characteristics.

  (1) The ratio of the block unit prediction method in the high resolution image is higher than the ratio of the block unit prediction method in the low resolution image.

  (2) The ratio of the block unit prediction method when the allocated code amount Ic per block is large is higher than the ratio of the block unit prediction method when Ic is small.

  As described above, the image coding apparatus according to the present embodiment estimates the amount of generated codes in all prediction methods by selecting an appropriate prediction method candidate according to control information based on the recording mode. This is not necessary, and the prediction method can be determined with a small amount of calculation.

  Also, since the encoded data stored in the recording medium 18 has the characteristics as described in (1) and (2) above, it is possible to select an appropriate prediction method according to the resolution of the image and the allocated code amount. The image quality after encoding can be improved.

  FIG. 3 is a block diagram showing the configuration of the image coding apparatus according to the second embodiment of the present invention.

  In FIG. 2, reference numeral 21 denotes a code amount control unit 21 that controls the generated code amount by measuring the generated code amount of each block and sequentially changing the encoding parameter such as the quantization width on the basis of the measurement result. It is. The other constituent elements are the same as those in the first embodiment, and thus the same reference numerals are given and description thereof is omitted.

  An encoding operation by the image encoding apparatus according to the present embodiment will be described with reference to the operation flow of FIG.

  When encoding is started, first, the prediction method control unit 11 determines a prediction method candidate to be used for the block (S1).

The prediction scheme candidate is input to the prediction scheme control unit 11 as control information by the code amount allocated to the block determined by the code amount control unit 21 or the quantization width used for the quantization of the block, and has a predetermined threshold value. It is determined by comparing with. The prediction method candidate determination method in the case of using the allocated code amount of the block as the control information is the same as in the first embodiment, and the description thereof is omitted. Here, the quantization width Iq and a predetermined threshold value TH4 are used. A case where a prediction method candidate is determined will be described. For example,
If Iq> TH4, all prediction methods in block units are set as prediction method candidates for the image.

  If Iq ≦ TH4, all prediction methods for each sub-block are set as prediction method candidates for the image.

  When the quantization width is large, the quality of the image generated by the encoding is lowered, so that the prediction accuracy is lowered and the difference in prediction efficiency between the prediction methods is reduced. In addition, the ratio of the code amount indicating the prediction method in the entire code amount is relatively high. Therefore, when the quantization width is large, it is possible to suppress the amount of code required for encoding the prediction scheme and increase the encoding efficiency by selecting a block-unit prediction scheme. On the other hand, when the quantization width is small, it is possible to increase the coding efficiency by selecting a prediction method in units of sub-blocks with relatively high prediction accuracy.

  Therefore, it is possible to select an appropriate prediction method candidate corresponding to the quantization width of the block by previously determining an appropriate threshold value TH4 through experiments. As in the case of the first embodiment, it is also possible to select a prediction method candidate step by step using a plurality of threshold values.

  Next, a prediction method to be used for the block to be encoded is determined from the selected prediction method candidates (S2). In addition, when using the prediction method per subblock, the prediction method used for every subblock is determined. Any method may be used as a method for determining a prediction method used in the block from the selected prediction method candidates. For example, a prediction method that estimates the generated code amount for each prediction method candidate and minimizes the generated code amount can be used.

  Note that this step can be omitted when only one type of prediction method is selected as a prediction method candidate.

  Next, the block is encoded based on the determined prediction method (S3). The predicted image generation unit 11 refers to the encoded image stored in the image storage unit 16 based on the designated prediction method, and generates a predicted image of the block or the subblock. The difference image generation unit 13 creates a difference image between the encoding target image of the block or the subblock and the prediction image, and the image encoding unit 14 determines the type of prediction method and the difference image of the block or the subblock. And the encoded data are sequentially stored in the recording medium 18.

  Note that the quantization width determined by the code amount control unit 21 is used as the quantization width in encoding. Further, the code amount control unit 21 measures the generated code amount of the block, and determines the quantization width used for the subsequent block encoding. The difference image encoded by the image encoding unit 14 is combined with the predicted image by the image combining unit 15 and stored in the image storage unit 16 as the encoded image of the block or the sub-block. Used for sub-block prediction.

  The operations of S1 to S3 are repeated for all the blocks of the encoding target image (S4), and the encoding of the image is completed.

  Thus, the encoded data stored in the recording medium 18 according to the present invention is such that the ratio of the block unit prediction method in the block with the large quantization width Iq is compared with the ratio of the block unit prediction method in the block with the small Iq. It has high characteristics.

  As described above, in the image coding apparatus according to the present embodiment, an appropriate prediction method candidate is selected for each block according to control information, and it is not necessary to estimate the amount of generated code in all prediction methods. The prediction method can be determined by the amount of calculation.

  Further, since the encoded data stored in the recording medium 18 has the above-described characteristics, an appropriate prediction method can be selected according to the quantization width for each block, and the image quality after encoding is improved. I can do it.

  FIG. 5 is a block diagram showing a configuration of an image encoding device according to the third embodiment of the present invention.

  In FIG. 3, 31 is a code amount estimation unit for estimating a code amount generated by encoding a difference image, and 32 is for selecting either the image encoding unit 14 or the code amount estimation unit 31 and outputting a difference image. Switch. The same components as those in the previous embodiment are given the same reference numerals, and the description thereof is omitted.

  An encoding operation in the image encoding apparatus of the present embodiment will be described below with reference to the operation flow of FIG.

  When encoding is started, first, the prediction method control unit 11 selects one predetermined prediction method from a plurality of prediction methods in block units, and the prediction image generation unit 12 uses the prediction method to select the block. A difference image generation unit 13 generates a difference image of the block, and a code amount estimation unit 31 obtains an estimated code amount Cb of the block in the prediction method.

  Similarly, the prediction method control unit 11 selects one predetermined prediction method from a plurality of prediction methods in units of sub-blocks, and the prediction image generation unit 12 predicts prediction images and differences for all sub-blocks constituting the block. The image generation unit 13 generates a difference image, and the code amount estimation unit 31 obtains an estimated code amount Cs of the block in the prediction method (S1).

  Note that an arbitrary prediction method can be selected as the prediction method used for the code amount estimation. However, the general prediction efficiency of each prediction method is determined in advance, and the prediction efficiency is generally the highest among the prediction methods for each block and sub-block. It is possible to increase the encoding efficiency by selecting a good prediction method.

  Moreover, the structure which uses a different prediction method for every block may be sufficient as the prediction method used for the said code amount estimation. For example, when encoding the image, it is possible to increase the encoding efficiency by measuring the frequency of use of each prediction method and using the prediction method having the highest use frequency at the time of encoding the block. The use frequency measurement result of the prediction method is input to the prediction method control unit 11 as control information.

Next, the prediction method control unit 11 determines a prediction method candidate used for encoding the block based on the code amount estimation result of the code amount estimation unit 17 and the predetermined threshold value TH5 (S2). For example,
When the result of the code amount estimation in S1 is Cb ≧ Cs + TH5, all prediction methods in units of sub-blocks are set as prediction method candidates in the block.

  When the result of the code amount estimation in S1 is Cb <Cs + TH5, all prediction methods for each block are set as prediction method candidates in the block.

  The threshold TH5 may be configured to be changeable according to I as a function TH5 = F (I) of the control information I given from the outside.

For example, by using a ratio Ib in which a block-based prediction method is used for encoding adjacent blocks as control information,
TH5 = F (Ib) = α1 · Ib + β1 (α1 and β1 are predetermined constants)
And In this case, when a block-based prediction method is frequently used in adjacent blocks, the value of TH5 is increased so that the block-based prediction method is easily selected, and a prediction method in sub-block units is set in adjacent blocks. When many are used, the value of TH5 is reduced so that the prediction method in units of sub-blocks can be easily selected.

  With such a configuration, it is possible to select an optimal prediction method with high accuracy using the correlation of images between adjacent blocks, and it is possible to improve coding efficiency. Note that F (Ib) is not limited to the linear function of Ib, but may be an arbitrary function obtained by obtaining a correlation between an adjacent block and the block based on an experiment in advance. For example, a function as shown in the graph of FIG. 7 may be used.

When using the quantization width Iq used for encoding the block as control information, for example,
TH5 = F (Iq) = α2 · Iq + β2 (α2 and β2 are predetermined constants)
And In this case, when the quantization width Iq is large, TH5 is large so that a block-based prediction method is easily selected, and when the quantization width Iq is small, a sub-block-based prediction method is easily selected. Thus, by adopting a configuration as described above in which the value of TH5 becomes smaller, it is possible to select an appropriate prediction method candidate according to the quantization width of the block, as in the second embodiment.

  Note that F (Iq) is not limited to the linear function of Iq, but may be an arbitrary function based on a predetermined experimental result, as in the case of F (Ib).

Further, using the allocated code amount Ic of the block as control information,
TH5 = F (Ic) = − α3 · Ic + β3 (α3 and β3 are predetermined constants)
As such, prediction candidates may be determined.

  Note that when any control information is used, the configuration may be such that prediction method candidates are switched in stages using a plurality of threshold values as in the first and second embodiments.

  Next, a prediction method used in the block is determined from the selected prediction method candidates (S3). In addition, when using the prediction method per subblock, the prediction method used from the said prediction method candidate is determined for every subblock. Any method may be used as a method for determining a prediction method used in the block from the selected prediction method candidates.

  For example, the generated code amount is estimated for each prediction method candidate in the same manner as the conventional method, and a prediction method that minimizes the generated code amount is used. Note that this step can be omitted when only one type of prediction method is selected as a prediction method candidate.

  Next, the block is encoded based on the determined prediction method (S4). The predicted image generation unit 11 refers to the encoded image stored in the image storage unit 16 based on the designated prediction method, and generates a predicted image of the block or the subblock. The difference image generation unit 13 creates a difference image between the image to be encoded of the block or the subblock and the prediction image, and the image encoding unit 14 determines the type of the prediction method and the difference image of the block or the subblock. The encoded data is sequentially stored in the recording medium 18. The encoded difference image is combined with the predicted image by the image combining unit 15 and stored in the image storage unit 16 as the encoded image of the block or the sub-block.

  The operations of S1 to S4 are repeated for all the blocks of the encoding target image (S5), and the encoding of the image is completed.

  Thus, the encoded data stored in the recording medium 18 according to the present invention has the following characteristics.

  (1) When the ratio Ib in which the block unit prediction method is used for adjacent blocks is high, the block unit prediction method ratio of the block is higher than the block prediction method ratio when Ib is low.

  (2) The ratio of the block unit prediction method in the block having a large quantization width Iq is higher than the ratio of the block unit prediction method in the block having a small Iq.

  As described above, in the image coding apparatus according to the present embodiment, all prediction schemes are selected in order to select prediction scheme candidates to be used in the block based on the code amount estimation by the prediction scheme representing each block unit and subblock unit. Therefore, it is possible to determine the prediction method with a small amount of calculation without requiring estimation of the generated code amount.

  Further, by using the control information, it is possible to select an appropriate prediction method candidate, and it is possible to efficiently determine a prediction method with high coding efficiency.

  Also, since the encoded data stored in the recording medium 18 has the characteristics as described in (1) and (2) above, appropriate prediction according to the correlation of the prediction method with adjacent blocks or the quantization width for each block. The method can be selected, and the image quality after encoding can be improved.

  FIG. 8 is a block diagram showing a configuration of an image encoding device according to the fourth embodiment of the present invention.

  In FIG. 8, reference numeral 41 denotes an image analysis unit for analyzing an input image. The same components as those in the previous embodiment are given the same reference numerals, and the description thereof is omitted.

  The operation of the image coding apparatus according to the present embodiment will be described below with reference to the operation flow of FIG.

  When encoding is started, first, the image analysis unit 21 performs image feature analysis in the block to be encoded (S1). The feature analysis is performed by evaluating the image similarity in each sub-block in the encoding target block.

  As the similarity S used for feature analysis, for example, for each sub-block,

  In this case, it is determined that the smaller the S, the smaller the variation between the sub-blocks and the higher the similarity, and the larger the S, the greater the variation between the sub-blocks and the lower the similarity.

  Alternatively, a method of selecting one predetermined sub-block, obtaining a difference between the image of the sub-block and an image of another sub-block in the block for each sub-block, and evaluating the similarity based on the sum of the differences. You may use.

  For example, when using the sum of the magnitudes of differences between the sub-block index k = 0 in FIG. 10 and the image in the other sub-blocks, the similarity S is expressed by the following equation.

  In this case, it is determined that the smaller the S, the smaller the error between the sub-blocks and the higher the similarity, and the larger the S, the greater the error between the sub-blocks and the lower the similarity.

Next, the prediction method control unit 11 determines a prediction method candidate used for encoding the block based on the result of the image feature analysis and predetermined threshold values TH6 and TH7 (where TH6 <TH7) (S2). For example,
If S <TH6 or TH7 ≦ S, all coding methods in units of blocks are set as prediction method candidates for the block.

  If TH6 ≦ S <TH7, all coding schemes in units of sub-blocks are set as prediction scheme candidates for the block.

  When the degree of similarity between sub-blocks is high (when S is small), high prediction accuracy can be obtained using either the block unit or sub-block unit prediction method, and either block unit or sub-block unit prediction method is used. However, the difference in coding efficiency of the difference image is small. Therefore, by selecting a prediction method in units of blocks, it is possible to suppress the amount of code indicating the prediction method and increase the coding efficiency.

  On the other hand, when the degree of similarity between sub-blocks is low (when S is large), the prediction accuracy is low even when using a prediction method in units of blocks or sub-blocks. Even if it is used, the difference in coding efficiency of the difference image is small. Therefore, by selecting a prediction method in units of blocks, it is possible to suppress the amount of code indicating the prediction method and increase the coding efficiency.

  In cases other than the above, it is possible to increase the coding efficiency by selecting a prediction method in units of sub-blocks with high prediction accuracy.

  As in the case of the first to third embodiments, the threshold values TH6 and TH7 may be changed according to the control information. Further, the threshold value is increased and predictions in block units and sub-block units occupying the prediction scheme candidates are possible. A configuration in which the ratio of the method is changed in stages may be used.

  Next, a prediction method used in the block is determined from the selected prediction method candidates (S3). In addition, when using the prediction method per subblock, the prediction method used for every subblock is determined. Any method may be used as a method for determining a prediction method used in the block from the selected prediction method candidates. For example, as in the conventional method, the generated code amount is estimated for all prediction method candidates, and a prediction method that minimizes the generated code amount is used.

  Note that this step can be omitted when only one type of prediction method is selected as a prediction method candidate.

  Next, the block is encoded based on the determined prediction method (S4). The predicted image generation unit 11 refers to the encoded image stored in the image storage unit 16 based on the designated prediction method, and generates a predicted image of the block or the subblock. The difference image generation unit 13 creates a difference image between the image to be encoded of the block or the subblock and the prediction image, and the image encoding unit 14 determines the type of the prediction method and the difference image of the block or the subblock. The encoded data is sequentially stored in the recording medium 18. The encoded difference image is combined with the predicted image by the image combining unit 15 and stored in the image storage unit 16 as the encoded image of the block or the sub-block.

  As in the first to third embodiments, when the threshold values TH6 and TH7 are changed according to the control information I, the encoded data stored in the recording medium 18 according to the present invention is the control information I. The ratio of the prediction method for each block changes according to the above.

  As described above, the image coding apparatus according to the present embodiment selects a prediction method candidate to be used in the block based on the result of image feature analysis, and therefore does not need to estimate a generated code amount in all prediction methods. The prediction method can be determined with a small amount of calculation.

  In addition, since the encoded data stored in the recording medium 18 has the above characteristics, an appropriate prediction method according to the control information can be selected, and the image quality after encoding can be improved.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a block diagram which shows the structure of the image coding apparatus in the 1st Embodiment of this invention. It is a figure which shows the operation | movement flow of the image coding apparatus of 1st Embodiment. It is a block diagram which shows the structure of the image coding apparatus in the 2nd Embodiment of this invention. It is a figure which shows the operation | movement flow of the image coding apparatus of 2nd Embodiment. It is a block diagram which shows the structure of the image coding apparatus in the 3rd Embodiment of this invention. It is a figure which shows the operation | movement flow of the image coding apparatus of 3rd Embodiment. It is a figure which shows the example of the function F (Ib). It is a block diagram which shows the structure of the image coding apparatus in the 4th Embodiment of this invention. It is a figure which shows the operation | movement flow of the image coding apparatus of 4th Embodiment. It is a figure which shows the outline | summary of a block and a subblock. It is a figure which shows the outline | summary of a prediction system. It is a figure which shows the operation | movement flow of the conventional image coding apparatus.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 11 Prediction system control part, 12 Prediction image generation part, 13 Difference image generation part, 14 Image encoding part, 15 Image composition part, 16 Image storage part, 17 Recording mode selection part, 18 Recording medium, 21 Code amount control part, 31 Code amount estimation unit, 32 changeover switch, 41 image analysis unit.

Claims (1)

In an image encoding device that performs intra-frame prediction encoding using a plurality of prediction schemes that can be changed for each block or for each sub-block obtained by further dividing the block, on a block-divided image,
A part of the plurality of prediction methods is set as a prediction method candidate under a predetermined condition, a prediction method of the block is determined from the prediction method candidate ,
The predetermined condition is an image compression rate. When the compression rate is high, the ratio of the prediction method for each block in the prediction method candidate is high, and when the compression rate is low, the prediction is performed. An image coding apparatus characterized by increasing a ratio of a prediction method for each sub-block in a method candidate .

Images