JP2010288181A - Moving image encoding method, moving image encoding apparatus, and moving image encoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To drastically reduce prediction errors and improve the encoding efficiency owing to error averaging, by selecting a signal with a minimum error evaluation value as a predicted image signal, among a plurality of motion compensation predicted image signals and one or more composite motion compensation predicted image signals. <P>SOLUTION: Using a motion vector, a motion compensation and image composition selector 111 applies motion compensation prediction to reference image signals up-sampled separately by filters 108-110. Then, the motion compensation and image composition selector 111 generates a motion compensation predicted image signal, respectively. Next, the motion compensation and image composition selector 111 synthesizes two pieces of motion compensation predicted image signal to generate synthetic image signal. Then, the motion compensation and image composition selector 111 adopts image signal with a minimum error evaluation value as a predicted image signal, among three pieces of motion compensation predicted image signal and three pieces of synthetic image signal. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a moving image encoding method, a moving image encoding device, and a moving image encoding program, and more particularly, to a moving image encoding method, a moving image encoding device, and a moving image encoding program equipped with motion compensation prediction processing.

  In motion picture compression coding represented by MPEG (Moving Picture Experts Group), motion compensation predictive coding that compresses a code amount using correlation between frames is often used. In motion compensation predictive coding, a motion vector indicating the relative positional relationship between each block of the coded image and the reference image is required. As for the motion vector, as the error evaluation value between blocks is smaller, the accuracy of motion compensation prediction is improved and the coding efficiency is improved.

  Here, the error evaluation value is generally a prediction error in which a difference between blocks is expressed by an absolute difference sum or a sum of squared differences using block matching, a motion vector code amount, DCT (discrete cosine transform: This is expressed by code amount conversion based on the coefficient code amount (discrete cosine transform).

  As techniques for improving the accuracy of motion compensation prediction, (1) bidirectional prediction using a B picture (bidirectional prediction picture) and (2) fractional accuracy motion compensation prediction are known.

  The bi-directional prediction using the B picture in (1) sets two reference images and combines two images that have been motion-compensated predicted using the motion vectors detected for each reference image. This is a technique for reducing the prediction error by averaging. However, two motion vectors are required for bidirectional prediction, and motion vector information is increased by one for unidirectional prediction. Therefore, when the increase in motion vector information is larger than the decrease in prediction error information, The encoding efficiency cannot be improved.

  On the other hand, the fractional accuracy motion compensated prediction of (2) above is actually performed when motion compensated prediction is performed by interpolating or interpolating the pixels of the reference image (this is referred to as “upsampling” in this specification). Up-sampling is performed using the pixel at the integer pixel position on the reference image calculated in step 1, and the pixel value at the 1/2 pixel position or 1/4 pixel position is calculated, and the motion vector is searched using the pixel value at the pixel position. By doing so, it is a technique that can perform motion compensation prediction with higher accuracy. Thereby, since the motion of the object in the video signal can be accurately expressed with fractional accuracy, the encoding efficiency can be further improved.

  The motion compensation prediction with fractional accuracy will be described in more detail. ISO (International Organization for Standardization) and IEC (International Electrotechnical Commission) standardize the first joint technical committee established as a method for upsampling reference images with fractional accuracy. According to MPEG-4 AVC (Advanced Video Coding) which is the latest moving picture compression coding standard, the pixel value with the fractional accuracy is obtained by filtering the reference picture with a 6-tap filter which is a linear filter.

  In this 6-tap filter, for example, a reference image signal whose motion vector is to be searched by block matching with a block of an encoding target image signal is represented by a pixel position with 1 pixel accuracy indicated by a circle in FIG. For example, the pixel value of a pixel with 1/2 pixel accuracy indicated by b at the middle position of each pixel with 1 pixel accuracy is arranged in the horizontal direction around the pixel b. The pixel values are generated based on the pixel values of six single-pixel precision pixels indicated by E, F, G, H, I, and J. On the other hand, the pixel value of the pixel with half-pixel accuracy indicated by h is six one-pixel accuracy indicated by A, C, G, M, R, and T arranged in the vertical direction around the pixel h. Generated based on the pixel value of the pixel. Similarly, the pixel value of each pixel such as aa, bb, cc, dd, ee, ff, gg, hh, b, m, s, etc. with 1/2 pixel accuracy is only in the vertical direction centered on the pixel or in the horizontal direction. It is generated from pixel values of six pixels with one pixel accuracy arranged only in the direction (generated using a 6-tap filter only in the vertical direction or only in the horizontal direction).

  On the other hand, the pixel value of a pixel with 1/2 pixel accuracy indicated by j in FIG. 11 is six pixels cc, dd, h, m, It is generated from pixel values of ee and ff (or pixel values of six pixels aa, bb, b, s, gg, and hh with 1/2 pixel accuracy arranged in the vertical direction). The pixel values of the six pixels cc, dd, h, m, ee, and ff for generating the pixel j are the pixel values of six pixels with one pixel accuracy in the vertical direction around the pixel, respectively. Therefore, after all, the pixel value of the pixel j is obtained by applying the 6-tap filter in both the vertical direction and the horizontal direction.

  Further, the pixel values of the pixels f, i, k, and q having the ¼ pixel accuracy are the above-described pixel j and the pixels b, h, It is generated from the pixel values of m and s. Accordingly, since the pixel values of these pixels f, i, k, and q also use the value of the pixel j, they are obtained by applying the 6-tap filter in both the vertical direction and the horizontal direction. The same applies to the pixels with ¼ pixel accuracy for which pixel values are obtained from other pixels with ½ pixel accuracy. When these are put together, each pixel having a pixel value calculated by a formula as shown in FIG. 12 is obtained by a 6-tap filter. In FIG. 11, the positions of pixels up-sampled by applying the 6-tap filter in both the vertical direction and the horizontal direction are indicated by crosses.

  An up-sampling method for generating a pixel value with 1/4 pixel accuracy using a 6-tap filter in which the filter coefficient as shown in FIG. 12 is fixed to (1, -5,20,20, -5,1) is as follows: Since the filter coefficient is fixed, the encoding device does not require the amount of calculation for determining the filter coefficient, and a circuit corresponding to one filtering method defined in both the encoding / decoding devices is prepared. Since it is only necessary to do this, the circuit scale is reduced.

  However, regardless of the state of the reference image, if all the reference images are upsampled with a fixed filter coefficient, the optimal upsampling cannot be performed if the reference image is degraded. As a result, the encoding efficiency decreases. This is because when a fixed filter coefficient is used, the filter is usually designed without considering the encoding deterioration. As a result, when motion compensation prediction is performed using a reference image including coding degradation, optimal upsampling cannot be performed.

  Therefore, in Patent Document 1 or the like, in order to perform optimum upsampling for each reference image, the filter coefficient is set so that the error evaluation value between the original image and the motion compensation predicted image is minimized for each reference image. Is determined.

  The determination of the filter coefficient that minimizes the error evaluation value between the original image and the motion compensation predicted image will be specifically described. The determination of the filter coefficient is a process performed on the encoding device side. First, a reference image is up-sampled using a filter having a fixed filter coefficient with respect to a reference image to be subjected to motion compensation prediction, and a motion vector is detected.

  Next, the coefficient of the upsampling filter is determined so that the error evaluation value between the original image to be encoded and the image subjected to motion compensation prediction using the detected motion vector becomes small. For example, the calculation is performed by the following formula using a non-separable two-dimensional Wiener filter.

ただし、上式中、Ｓx,yは原画像を表し、Ｐx,yは参照画像を表す。また、(ｅSP)2は誤差評価値（ここでは差分二乗和を用いる）であり、(ｅSP)2が最小となるようにフィルタ係数ｈi,jSPを決定する。

In the above equation, Sx, y represents the original image, and Px, y represents the reference image. Further, (eSP) 2 is an error evaluation value (here, the sum of squared differences is used), and the filter coefficient hi, jSP is determined so that (eSP) 2 is minimized.

  Various methods have been proposed for the filter coefficient determination algorithm represented by the above, and it is preferable to appropriately determine which method is to be adopted in consideration of the balance between the encoding efficiency and the calculation amount. This is because there is a trade-off relationship between the encoding efficiency and the calculation amount.

  When the filter coefficient of the 6-tap filter is calculated as (2, -4,18,18, -4,2) using a predetermined filter coefficient determination algorithm, the pixel of the reference image is calculated using this 6-tap filter. Upsampling improves the coding efficiency of the entire picture. However, locally (for example, in block units), the calculated filter coefficients do not necessarily have better encoding efficiency than other filter coefficients.

  Thus, as in Patent Document 2, a technique for further improving the coding efficiency by preparing a plurality of filtering methods and adaptively selecting filter coefficients in units of blocks has been proposed.

Special table 2008-536414 gazette Special table 2008-544708 gazette

  As described above, in the conventional moving image coding apparatus, when generating a motion compensated prediction image with dispersion accuracy, the coefficient of the upsampling filter is determined so that an error from the original image becomes small. As a result, the prediction error to be transmitted is reduced and the coding efficiency is improved. Furthermore, it is possible to further improve the coding efficiency by preparing a plurality of filter coefficients and adaptively selecting the filter coefficients in a predetermined unit such as a block unit.

  However, if only a filter coefficient is adaptively selected from such a plurality of filter coefficients, two motion vectors are required when bi-directional prediction is used to further reduce the prediction error using error averaging. There is no change in anything. Therefore, when the amount of increase in motion vector information is larger than the amount of decrease in prediction error information due to bidirectional prediction, the encoding efficiency cannot be improved. Further, when only one reference image can be used as in a P picture (forward prediction picture), there is no means for further reducing the prediction error by using error averaging.

  The present invention has been made in view of the above points. Among the plurality of motion compensation prediction image signals and one or more combined motion compensation prediction image signals generated by combining two or more motion compensation prediction image signals, an error is generated. A moving picture coding method, a moving picture coding apparatus, and a moving picture that select a signal with the smallest evaluation value and use it as a predicted picture signal to significantly reduce the prediction error and improve coding efficiency by averaging the errors. An object is to provide an encoding program.

  In order to achieve the above object, the moving image encoding method of the present invention generates a reference image signal by locally decoding an encoded signal obtained by encoding an encoding target image signal in an input moving image signal, In a moving picture coding method for performing motion compensated prediction coding of a current coding target image signal in an input moving picture signal using a predicted picture signal obtained by motion compensation prediction of the reference picture signal, the input moving picture signal Each of the filter processing by the first step of detecting the motion vector of the current encoding target image signal in FIG. 1, two or more filter processing from the generated reference image signal, and motion compensation processing based on the same motion vector A second step of generating a motion compensated prediction image signal corresponding to each of the motion compensation prediction image signals, and two or more motion compensations of different combinations among the generated motion compensation prediction image signals. A third step of synthesizing the predicted image signals to generate one or more combined motion compensated predicted image signals; and a signal having the smallest error evaluation value among the plurality of motion compensated predicted image signals and the combined motion compensated predicted image signals. A fourth step that is selected and output as a predicted image signal, and a difference signal between the current encoding target image signal and the predicted image signal selected in the fourth step is subjected to orthogonal transform and quantization to be an encoded signal And a reference image signal for specifying the generated reference image signal, the same motion vector, and the filter processing used to generate the selected predicted image signal, respectively, Specific information, motion vector specifying information, and filtering processing specifying information are added to a moving image encoded signal generated by performing a predetermined encoding process on the quantized signal and output in a sixth step. And a seventh step of generating a new reference image signal by local decoding processing by adding the selected predicted image signal to a signal obtained by inverse quantization and inverse orthogonal transform of the quantized signal. It is characterized by including.

  Here, in the third step, one or more synthesis is performed by weighting and synthesizing each of two or more motion compensation prediction image signals of different combinations among the plurality of motion compensation prediction image signals at a set synthesis ratio. A motion compensated predicted image signal is generated, and the sixth step is used to generate a predicted image signal when the fourth step selects a synthesized motion compensated predicted image signal and outputs it as a predicted image signal. The synthesis ratio information indicating the synthesized ratio may be further added to the moving image encoded signal.

  In order to achieve the above object, the moving image encoding apparatus of the present invention generates a reference image signal by locally decoding an encoded signal obtained by encoding an encoding target image signal in an input moving image signal. In the moving picture coding apparatus that performs motion compensated prediction coding of the current coding target image signal in the input moving picture signal using the predicted picture signal obtained by motion compensation prediction of the reference picture signal, A motion vector detecting means for detecting a motion vector of the current encoding target image signal in the image signal, two or more filtering processes from the generated reference image signal, and a motion compensation process based on the same motion vector, A motion compensated predicted image signal generating unit that generates a motion compensated predicted image signal corresponding to each of the filtering processes and a plurality of generated motion compensated predicted image signals are different from each other. A combined motion compensated predicted image signal generating means for generating one or more combined motion compensated predicted image signals by combining two or more motion compensated predicted image signals in combination, a plurality of motion compensated predicted image signals and a combined motion compensated predicted image A difference between a prediction image signal selection unit that selects a signal having the smallest error evaluation value from the signals and outputs the selected signal as a prediction image signal, and a prediction image signal selected by the current encoding target image signal and the prediction image signal selection unit The quantized signal generating means for generating a quantized signal that is an encoded signal by orthogonal transform and quantization with respect to the signal, and the generated reference image signal used for generating the selected predicted image signal are the same The reference image signal specifying information, the motion vector specifying information, and the filter processing specifying information for specifying the motion vector and the filtering process are respectively added to the quantized signal by a predetermined code. Adds the selected predicted image signal to the signal obtained by dequantizing and inverse orthogonal transforming the quantized signal and adding the output to the encoded video signal generated by the quantization process And a reference image signal generating means for generating a new reference image signal by local decoding processing.

  Here, the synthesized motion compensated prediction image signal generation unit weights and synthesizes each of two or more motion compensation prediction image signals of different combinations among the plurality of motion compensation prediction image signals at a set composition ratio. One or more synthesized motion compensated predicted image signals are generated, and the encoding means selects the predicted image signal when the predicted image signal selecting means selects the synthesized motion compensated predicted image signal and outputs it as a predicted image signal. The synthesis ratio information indicating the synthesis ratio used for generating the image may be further added to the encoded video image signal.

Furthermore, in order to achieve the above object, the moving picture coding program of the present invention generates a reference picture signal by locally decoding the coded signal obtained by coding the coding target picture signal in the input moving picture signal. Then, using a prediction image signal obtained by motion compensation prediction of the reference image signal, a moving image encoding process for performing motion compensation prediction encoding of the current encoding target image signal in the input moving image signal is executed on the computer. In the moving image encoding program to be
A first step of detecting a motion vector of a current encoding target image signal in an input moving image signal in the computer, and any one of a plurality of filters having different filter coefficients from a generated reference image signal A second step of generating a plurality of motion compensated predicted image signals by a filter process using the same motion vector and a motion compensation process based on the same motion vector, and combinations of mutually different combinations among the plurality of motion compensated predicted image signals A third step of synthesizing two or more motion-compensated predicted image signals to generate one or more synthesized motion-compensated predicted image signals, and an error evaluation value among a plurality of motion-compensated predicted image signals and synthesized motion-compensated predicted image signals Is selected by the fourth step of selecting the smallest signal and outputting it as the predicted image signal, and the current encoding target image signal and the fourth step. A fifth step for generating a quantized signal, which is an encoded signal, by performing orthogonal transform and quantization on the difference signal from the measured image signal, and the generated reference used to generate the selected predicted image signal A moving image generated by performing a predetermined encoding process on the quantized signal, the image signal, the same motion vector, and the reference image signal specifying information for specifying the filter, the motion vector specifying information, and the filter specifying information. A sixth step of adding and outputting to the encoded signal, and a local decoding process by adding the selected predicted image signal to the signal obtained by inverse quantization and inverse orthogonal transform of the quantized signal, And a seventh step of generating a new reference image signal.

  Here, in the third step of the moving picture coding program of the present invention, the combination ratio in which each of two or more motion compensated predicted image signals of different combinations among the plurality of motion compensated predicted image signals is set. Are combined by weighting to generate one or more combined motion compensated predicted image signals. In the sixth step, the fourth step selects the combined motion compensated predicted image signal and outputs it as a predicted image signal. At this time, synthesis ratio information indicating a synthesis ratio used for generating the predicted image signal may be further added to the moving image encoded signal.

  According to the present invention, a signal having a minimum error evaluation value is selected from among a plurality of motion compensation prediction image signals and one or more combined motion compensation prediction image signals generated by combining two or more motion compensation prediction image signals. By selecting it as a predicted image signal, it is possible to greatly reduce the prediction error and improve the coding efficiency by averaging the errors.

  Also, according to the present invention, since only one motion vector is required for generating a plurality of motion compensated prediction image signals, a highly accurate prediction image can be generated without an increase in motion vector information.

It is a block diagram of 1st Embodiment of the moving image encoder of this invention. 2 is a flowchart for explaining the operation of a motion compensation / image composition selection unit 111 in FIG. 1. It is explanatory drawing of an example of the synthetic | combination method of an image. It is explanatory drawing of an example of the image composition mode by the motion compensation and image composition selection part 111 in FIG. It is a figure which shows an example of the syntax of image composition mode information. It is a figure which shows an example of the semantics of mix_mode in FIG. It is a figure which shows an example of the syntax of weighted prediction. It is a figure which shows an example of the semantics of weight_idx in FIG. It is a block diagram of 2nd Embodiment of the moving image encoder of this invention. It is a block diagram of an example of a moving image decoding apparatus. It is explanatory drawing (the 1) of the up-sampling filter used by MPEG-4AVC. It is explanatory drawing (the 2) of the up sample filter used by MPEG-4AVC.

  Next, embodiments of the present invention will be described in detail with reference to the drawings. The present invention generates a plurality of motion compensated prediction images using a plurality of upsampling techniques for a reference image at the time of motion compensation prediction, and converts the generated motion compensated prediction image into a prediction image. This improves the encoding efficiency.

(First embodiment)
FIG. 1 shows a block diagram of a first embodiment of a video encoding apparatus according to the present invention.

  The moving picture coding apparatus 100 according to the present embodiment includes a local decoding unit 102, a subtractor 103, an orthogonal transformation / quantization unit 104, a motion vector detection unit 112, and an entropy coding unit 113.

  The local decoding unit 102 includes an inverse quantization / inverse orthogonal transform unit 105, an adder 106, a reference image memory 107, three filters 108, 109, and 110, and a motion compensation / image synthesis selection unit 111. A predicted image signal obtained by local decoding from the motion compensation / image synthesis selection unit 111 is output to the subtractor 103. In addition, the motion compensation / image synthesis selection unit 111 supplies the image synthesis mode information to the entropy encoding unit 113.

  The subtracter 103 subtracts an input image signal (input moving image signal) that is a moving image signal from the input terminal 101 and a predicted image signal from a motion compensation / image synthesis selection unit 111 described later, and uses the difference signal therebetween. A certain prediction error signal is generated. The orthogonal transform / quantization unit 104 performs orthogonal transform / quantization on the prediction error signal from the subtractor 103 to generate a quantized signal.

  The inverse quantization / inverse orthogonal transform unit 105 performs inverse quantization / inverse orthogonal transform on the quantized signal output from the orthogonal transform / quantization unit 104. The adder 106 adds the inversely quantized / inverse orthogonal transformed signal output from the inverse quantization / inverse orthogonal transform unit 105 and the predicted image signal from the motion compensation / image synthesis selection unit 111, and A decoded reference image signal is generated. The reference image memory 107 accumulates the locally decoded reference image signal.

  The filters 108, 109, and 110 filter the same reference image signal read from the reference image memory 107 with different filter coefficients, and upsample the reference image signal.

  Here, an example of a filtering method used in each of the filters 108, 109, and 110 will be introduced. Since the output signal of the filter 108 is also used for upsampling in the motion vector detection unit 112, it is desirable to perform filtering of fixed filter coefficients. For example, it is assumed that the filter 108 is a 6-tap filter (filter coefficient; 1, -5, 20, 20, -5, 1) used in MPEG-4 AVC.

  On the other hand, the filter coefficients of the filter 109 and the filter 110 are switched in units of pictures. For example, the filter coefficient of the filter 109 is set so that the error between the original image (input image signal) of the encoding target picture and the decoded image (reference image signal locally decoded by the local decoding unit 102) is minimized. It is determined using a Wiener filter that performs filtering of the arithmetic expression shown in Equation 1. However, calculating the optimal filter coefficient for the picture to be encoded requires performing motion compensation prediction for all the blocks in the picture once, determining the filter coefficient, and then selecting the optimal filter means for each block again. Therefore, it becomes substantially two-pass encoding. Therefore, the filter 109 is not used when 1-pass encoding is required.

  The filter 110 uses the filter coefficient calculated by using the Wiener filter in the filter 109 for the previous encoding target picture as it is. Using the filter coefficient calculated for the previous encoding target picture as it is can be performed by one-pass encoding since the filter coefficient has already been determined, and can also reduce the code amount of filter coefficient transmission. Is possible. However, the filtering method described here is only an example, and even if a filter coefficient generated by another method is used, it is included in the present invention.

  The motion vector detection unit 112 calculates a current encoding target in the input image signal from the input image signal and a signal obtained by up-sampling the decoded reference image signal stored in the reference image memory 107 using the filter 108. A motion vector of the image signal is detected.

  The motion compensation / image synthesis selection unit 111 uses the motion vector detected by the motion vector detection unit 112 to perform motion compensation prediction on each image signal obtained by up-sampling the reference image signal by the filters 108 to 110. I do. Also, the motion compensation / image synthesis selection unit 111 generates a predicted image signal with little prediction error from the original image signal by selecting or synthesizing each motion compensated predicted image signal. A method for generating the predicted image signal will be described in detail later.

  The entropy encoding unit 113 entropy-encodes the image signal orthogonally transformed and quantized by the orthogonal transform / quantization unit 104 and outputs a bit stream of the moving image encoded signal. The entropy encoding unit 113 also includes information indicating the upsampling method of the reference image signal used in the motion compensation / image synthesis selection unit 111, information indicating how the upsampled image signal is combined, and the like. The additional information (image synthesis mode information) is entropy-encoded and added to the moving image encoded signal. The entropy encoding unit 113 outputs a bit stream including a moving image encoded signal obtained by entropy encoding and image synthesis mode information to the outside of the apparatus via the output terminal 114.

  Here, as described later, the image synthesis mode information includes reference image signal specifying information for specifying the reference image signal used to generate the predicted image signal selected by the motion compensated image synthesis selecting unit 111, and a filter. Motion vector specification information for specifying the same motion vector used in 108 to 110, and filter processing specification for specifying the filter processing for generating the reference image signal used to generate the predicted image signal selected from the filters 108 to 110 Contains information.

  As described above, the moving image coding apparatus 100 according to the present embodiment prepares three filters for upsampling the reference image signal, the three motion compensated prediction image signals obtained thereby, and the three motion compensated predictions. It is characterized in that a predetermined one predicted image signal is selected from a plurality of motion compensated predicted image signals obtained by synthesizing images based on the image signal and used for the predicted image signal.

  Next, the motion compensation / image composition selection unit 111 as the main part of the present embodiment will be described in detail with reference to the flowchart of FIG. The flowchart of FIG. 2 is executed in units of motion compensation blocks.

  First, the motion compensation / image synthesis selection unit 111 performs motion compensation prediction on the reference image signals individually upsampled by the filters 108, 109, and 110 using the motion vector detected by the motion vector detection unit 112. To generate a motion compensated prediction image signal (step S11). Next, the motion compensation / image synthesis selection unit 111 performs upsampling with two of the three filters 108, 109, and 110 (here, the filters 108 and 109, the filters 109 and 110, and the filters 108 and 110). The two motion compensated prediction image signals obtained in step S11 are synthesized from the reference image signal thus generated to generate a synthesized image signal (step S12).

  The most common method for synthesizing image signals is to use an average value of two image signals. FIG. 3 schematically shows a state of image composition when the average value is used. In FIG. 3, the reference image block 161 corresponding to the encoding target block 151 in the encoding target image 150 that is the input image signal of the uplink reference image 160 of the first filter of one of the two filters described above. A motion compensation predicted reference image block indicated by the motion vector detected by the motion vector detection unit 112 is indicated by 162. On the other hand, with respect to the reference image block 171 corresponding to the encoding target block 151 in the encoding target image 150 of the input image signal of the uplink reference image 170 of the other second filter of the above two filters, A motion compensation predicted reference image block indicated by the same motion vector as described above is indicated by 172.

  In this case, the motion compensation / image synthesis selection unit 111 divides the sum of the two motion compensation predicted reference image blocks 162 and 172 by 2 to calculate a synthesized image signal that is an average value of them. Alternatively, when the average value is calculated with higher accuracy, 1 is added to the sum of the reference image blocks 162 and 172 and then divided by 2. This is an operation equivalent to rounding off. Of the image synthesis methods, a synthesis method (weighted prediction) other than the average value will be described later.

  Finally, the motion compensation / image synthesis selection unit 111 generates the three motion compensated prediction image signals obtained from the reference image signals separately upsampled by the three filters 108, 109, and 110, respectively, and generated in step S12. Of the three synthesized image signals, the image signal having the minimum error evaluation value is adopted as the predicted image signal (step S13).

  As described above, the motion compensation / image synthesis selection unit 111 selects which motion compensation out of the three motion compensation predicted image signals obtained from the reference image signals separately up-sampled by the three filters 108, 109, and 110, respectively. It is determined whether a predicted image signal is selected or which two motion compensated predicted image signals are selected from the three motion compensated predicted image signals and synthesized (image synthesis mode).

  FIG. 4 is an explanatory diagram of the image composition mode of the present embodiment. In the image synthesis mode according to the present embodiment, the motion compensated prediction image signal generated from the reference image signal upsampled by one of the three filters 108, 109, and 110 may be selected independently. , There are three ways as shown in S23. The case of selecting an image signal obtained by synthesizing two motion compensated prediction image signals generated from reference image signals upsampled by two of the three filters 108, 109, and 110 is indicated by S24, S25, and S26. There are three ways. Therefore, in the image composition mode of the present embodiment, one of the above six options is selected.

  The number of choices for this image synthesis mode depends on how many motion compensated prediction image signals output from the filter are used. For example, when there are two simplest filters, the number of image synthesis mode options is three, that is, two independent motion compensated prediction image signals and one composite image signal.

  Since it cannot be decoded unless the moving image decoding apparatus identifies which signal has been selected and combined in the image combining mode, the image combining mode information is entropy encoded by the entropy encoding unit 113 in FIG. It is transmitted as information in the bitstream.

  Next, a method for transmitting image composition mode information will be described.

  FIG. 5 shows an example of the syntax of the image composition mode. First, a flag (use_adaptive_filter_flag) indicating whether or not to use an adaptive filter is sent in units of 16-line width slices when a screen of one picture is sliced by 16 lines. However, this division method in which 16 lines are one slice is merely an example, and one picture may be one slice. Turn this flag on if you want to use a filter other than the default. In the case of the present embodiment, the filter 108 is described as a default filter, and the filters 109 and 110 are described as filters that define themselves.

  Next, as shown in FIG. 5, the number of filters (num_of_filter) defined other than the default is sent in units of slices. In the present embodiment, the number of filters (num_of_filter) is “2”. Then, the filter coefficient is sent for each filter. This filter coefficient transmission method does not have to send all the coefficients by utilizing the symmetry of the filter.

  A slice having a width of 16 lines is composed of a plurality of macroblocks having a width of 16 lines. Therefore, a macroblock unit (or sub-macroblock unit) composed of a luminance block of 16 pixels in the horizontal direction and 16 pixels in the vertical direction and two color difference blocks of 8 pixels in the horizontal direction and 8 pixels in the vertical direction, respectively, is shown in FIG. The image composition mode (mix_mode) is sent as follows. This image composition mode (mix_mode) is sent together with a reference image number (ref_idx) and a motion vector difference (mvd).

  FIG. 6 shows an example of the semantics of the image composition mode (mix_mode). In FIG. 6, the semantics indicating the meaning of the syntax are shown in FIG. 6, when the value is “0”, the used filter is filter # 1, when the value is “1”, the used filter is filter # 2, and when the value is “2” Is filter # 3, when the value is “4”, the filters used are filters # 1 and # 2, when the value is “5”, the filters used are filters # 2 and # 3, and when the value is “6”, the filters used are the filters Indicates # 1 and # 3. Here, the filters # 1, # 2, and # 3 are the filters 108, 109, and 110 in FIG. 1, respectively.

  However, FIG. 6 is only an example in the case of a P picture. Originally, an event most likely to be selected should be assigned to a small number. This assignment may be fixed, but the order may be changed by a slice header or the like.

  Moreover, the syntax in the case of transmitting this image composition mode information is only an example. For example, a flag to be transmitted in units of slices may be transmitted in units of pictures.

  As described above, the moving image encoding apparatus 100 according to the present embodiment generates a plurality of motion compensated prediction image signals by using a plurality of upsampling techniques for a reference image signal in motion compensation prediction. Predictive image signal is greatly increased by averaging one error signal from the selected motion compensation predicted image signal and a plurality of synthesized image signals obtained by combining multiple motion compensated predicted image signals. Thus, the coding efficiency can be improved.

  That is, even in a P picture that can only use unidirectional prediction in the past, an image signal obtained by synthesizing a plurality of motion compensation predicted image signals corresponding to bidirectional prediction of a B picture can be used as a predicted image signal. it can. As a result, the prediction error is reduced and the coding efficiency is greatly improved as compared with the case where synthesis is not used.

  In addition, in the conventional bi-prediction of a B picture, two motion vectors are required to generate a plurality of motion-compensated prediction images, and the amount of motion vector information increases as compared with uni-directional prediction. On the other hand, in the moving picture encoding apparatus 100 according to the present embodiment, in a method of generating a plurality of motion compensated prediction image signals using three filters and selecting one prediction image signal from them, a motion vector is Since only one is needed, the amount of motion vector information does not increase. Therefore, the problem that the prediction error becomes small but the code amount of the motion vector increases more than the conventional B picture bi-directional prediction is solved, and the prediction image signal with high accuracy without increasing the motion vector information is solved. Can be generated.

Here, since the upsampling filters 108, 109, and 110 used in this embodiment are linear filters, it is also possible to represent and transmit a composite image of a plurality of filters as a single filter. There are the following disadvantages.
(1) It is necessary to transmit the coefficient of a filter for generating a composite image, and the amount of codes increases.
(2) An integer filter having a limited bit depth cannot accurately represent a filter that generates the same image as the synthesized image of the present embodiment.
(3) A filter coefficient calculation calculation amount for expressing as a single filter is required.
(4) In order to cope with weighted prediction described later, it is necessary to calculate / transmit filter coefficients corresponding to the respective weighting coefficients.

  Therefore, it is clear that the composite image generation method of the present embodiment is superior to the case where a composite image of a plurality of filters is generated by a single filter.

  Next, another example of an image composition method will be described. In the description of the above embodiment, an example in which an average value of two motion compensated prediction image signals is used as an image synthesis method is shown, but the image synthesis method is not limited to this, and is used in MPEG-4 AVC. As in the weighted prediction, a combination of two motion compensated prediction image signals at a predetermined ratio may be used as the prediction image signal. In this case, a more efficient predicted image signal can be generated.

  FIG. 7 shows an example of the syntax of weighted prediction. First, as shown in FIG. 7, a flag (adaptive_filter_weighted_flag) indicating whether or not to perform weighted prediction in units of slices is included in the bitstream and sent. When weighted prediction is used, the accuracy of weighted prediction (weighted_accuracy_idx) is designated. Here, it is assumed that the accuracy when weighted_accuracy_idx = 0 (corresponding to the denominator at the time of image synthesis calculation) is “3”, and the accuracy when weighted_accuracy_idx = 1 is “4”.

  In a macroblock unit (or submacroblock unit), as shown in FIG. 7, a weighting index (weighted_idx) defined for each weighted prediction accuracy (weighted_accuracy_idx) is sent.

  FIG. 8 shows an example of the semantics of the weighted index (weighted_idx). When weighted_accuracy_idx = 0, weighted_idx = 0 means that images are synthesized with the ratio of the first filtered image being “2” and the ratio of the second filtered image being “1”. For example, the ratio of the filtering image of the filter 108 is “2” and the ratio of the filtering image of the filter 109 is “1”. Weighted_idx = 1 means that the image is synthesized with the ratio of the first filtered image being “1” and the ratio of the second filtered image being “2”.

Further, when weighted_accuracy_idx = 1, weighted_idx = 0 means that images are combined with the ratio of the first filtered image being “2” and the ratio of the second filtered image being “2”. For example, the ratio of the filtering image of the filter 108 is “2”, and the ratio of the filtering image of the filter 109 is “2”. Weighted_accuracy_idx = 1, weighted_idx = 1 means that the ratio of the first filtered image is “3”, and the ratio of the second filtered image is “1”, and the images are combined.
weighted_idx = 2 means that the ratio of the first filtered image is “1” and the ratio of the second filtered image is “3” to synthesize the images.

  This weighting ratio can be determined by various methods, but multiple weighting ratio candidates are prepared, images are actually combined based on the weighting ratios of each candidate, and the error between the combined image and the original image An algorithm that employs a weighting ratio with the smallest evaluation value has the highest coding efficiency.

(Second Embodiment)
FIG. 9 shows a block diagram of a second embodiment of the moving picture coding apparatus according to the present invention. In the figure, the same components as those in FIG. The moving picture coding apparatus 200 according to the present embodiment includes a local decoding unit 201, a subtractor 103, an orthogonal transformation / quantization unit 104, a motion vector detection unit 112, and an entropy coding unit 113. The configuration of the decoding unit 201 is different from the local decoding unit 102 of FIG.

  The local decoding unit 201 includes an inverse quantization / inverse orthogonal transform unit 105, an adder 106, a reference image memory 107, a filter 108, a motion compensation unit 202, filters 203 and 204, and an image synthesis / selection unit 205. The prediction image signal obtained by local decoding from the image synthesis / selection unit 205 is output to the subtractor 103. The image composition / selection unit 205 also supplies the image composition mode information to the entropy coding unit 113.

  The motion compensation unit 202 uses the motion vector detected by the motion vector detection unit 112 to perform motion compensation prediction on the image signal obtained by upsampling the reference image signal by the filter 108.

  The filters 203 and 204 have different filter coefficients, and individually filter the images subjected to motion compensation by the motion compensation unit 202. The filters 203 and 204 are not upsampling filters such as the filter 108, and perform filtering in which the output image signal has the same number of samples as the input image signal.

  The image synthesis / selection unit 205 includes two motion compensation predicted image signals filtered by the filters 203 and 204, and one synthesized image signal obtained by synthesizing the two motion compensation predicted image signals by a predetermined method. An image signal having a minimum error evaluation value (one image signal with a small prediction error from the input image signal) is selected as a predicted image signal. The predicted image signal generation method is the same as the generation method described in the first embodiment, and a description thereof will be omitted.

  The subtracter 103 generates a prediction error signal by subtracting the input image signal and the predicted image signal selected and output by the image synthesis / selection unit 205 and supplies the prediction error signal to the orthogonal transform / quantization unit 104.

  The difference from the moving image encoding apparatus 100 of the first embodiment is that the moving image encoding apparatus 100 up-samples the reference image signal with three filters 108 to 110 having different filter coefficients. While the obtained three motion-compensated prediction image signals after upsampling are selected and synthesized, the moving picture coding apparatus 200 according to the present embodiment selects one motion-compensated prediction image signal after up-sampling. That is, the two image signals filtered by the filters 203 and 204 are selected and combined.

  However, the moving image encoding apparatus 200 according to the present embodiment also generates a plurality of motion compensated prediction image signals, similarly to the moving image encoding apparatus 100 according to the first embodiment, and generates the plurality of motion compensated prediction images. Since the image signal having the minimum error evaluation value is selected from the signal and the composite image signal obtained by synthesizing a plurality of motion compensated prediction image signals, the prediction error is greatly reduced due to the error averaging. Efficiency can be improved.

  In addition, the moving picture coding apparatus 200 according to the present embodiment also solves the problem of conventional B-picture bi-prediction in which the prediction error is small but the code amount of the motion vector is further increased. A highly accurate predicted image signal can be generated without an increase in information.

  Next, a moving picture decoding apparatus for decoding a bit stream obtained by encoding with the moving picture encoding apparatus of the present invention will be described.

  FIG. 10 is a block diagram illustrating an example of a moving picture decoding apparatus. As illustrated in FIG. 10, the moving picture decoding apparatus 300 includes an entropy decoding unit 302 that performs entropy decoding on a bitstream input via an input terminal 301, a motion compensation / image synthesis unit 303, and an inverse quantization. An inverse orthogonal transform unit 304, an adder 305, a reference image memory 306, three upsampling filters 307, 308 and 309, and an output terminal 310.

  A bit stream generated by the video encoding device 100 in FIG. 1 or a bit stream obtained by causing a computer to execute the operation of each component of the video encoding device 100 is input to the input terminal 301. .

  The entropy decoding unit 302 entropy-decodes the bitstream input to the input terminal 301 to decode an orthogonally transformed / quantized image signal, and adds additional information such as a motion vector and image synthesis mode information. Decrypt.

  The motion compensation / image synthesis unit 303 performs motion compensation prediction on the upsampled reference image signals respectively supplied from the filters 307 to 309 using the motion vector supplied from the entropy decoding unit 302. A plurality of motion compensated prediction image signals are generated, and further, based on the image composition mode information supplied from the entropy decoding unit 302, image synthesis corresponding to the image composition mode is performed using the plurality of motion compensation prediction image signals. I do. This image synthesizing method is the same as the image synthesizing method in the above-described moving image encoding apparatus.

  The inverse quantization / inverse orthogonal transform unit 304 performs inverse quantization / inverse orthogonal transform on the orthogonally transformed / quantized image signal output from the entropy decoding unit 302 to decode a differential signal (prediction error signal). . The adder 305 adds the prediction error signal output from the inverse quantization / inverse orthogonal transform unit 304 and the motion compensated prediction image signal output from the motion compensation / image synthesis unit 303 to decode the image signal. The data is output to the output terminal 310 and also output to the reference image memory 306. The reference image memory 306 stores the decoded image signal supplied from the adder 305 as a reference image signal.

  The filters 307, 308, and 309 separately filter the same reference image signal supplied from the reference image memory 306 using different filter coefficients to generate an upsampled predicted image signal, The predicted image signal is output to the motion compensation / image synthesis unit 303. These filters 307, 308 and 309 are filters corresponding to the filters 108, 109 and 110 in FIG.

  The moving image decoding apparatus 300 prepares three filters (filters 307 to 309) for upsampling the reference image signal, and the motion compensation / image synthesis unit 303 performs upsampling on these three filters 307 to 309, respectively. The image signal generated according to the image synthesis mode information from the entropy decoding unit 302 based on the three reference image signals is characterized in that it is used as a predicted image.

  Next, details of the motion compensation / image synthesis unit 303, which is a main part of the video decoding device 300, will be described.

  The motion compensation / image synthesis unit 303 first uses the single motion vector supplied from the entropy decoding unit 302 for the upsampled reference image signal supplied from the filters 307 to 309 to perform motion compensation prediction. Processing is performed to generate three motion compensated prediction image signals.

  Subsequently, the motion compensation / image synthesis unit 303, when the image synthesis mode information supplied from the entropy decoding unit 302 indicates that two predetermined motion compensation predicted image signals are synthesized, Two motion compensation predicted image signals designated in the image synthesis mode are synthesized from the motion compensated prediction image signal. This image composition method is generally the method of calculating the average value of two image signals described with reference to FIG. 3, but may be the weighted image composition described with reference to FIGS. In any case, the image composition method indicated by the image composition mode information is adopted as the image composition method.

  In addition, the motion compensation / image synthesis unit 303 selects a predetermined motion compensation predicted image signal from among the three motion compensation predicted image signals as the image synthesis mode information supplied from the entropy decoding unit 302. , One motion compensated predicted image signal designated in the image synthesis mode is selected from the three motion compensated predicted image signals and output.

  In this way, the motion compensation / image synthesis unit 302 supplies the motion compensated predicted image signal obtained by selection or synthesis to the adder 305 as a predicted image signal. The adder 305 adds the predicted image signal and the decoded prediction error signal from the inverse quantization / inverse orthogonal transform unit 304, and outputs a decoded image signal.

  According to the moving picture decoding apparatus 300, a plurality of motion compensated prediction image signals are generated by using a plurality of upsampling techniques with respect to a reference image signal at the time of motion compensation prediction, and the generated plurality of motion compensated prediction images are generated. Prediction error due to error averaging by making the image signal specified by the image synthesis mode information out of a plurality of synthesized image signals obtained by combining the signal and a plurality of motion compensated predicted image signals into a predicted image signal Thus, the coding efficiency can be improved.

  That is, even in a P picture that can only use unidirectional prediction in the past, an image signal obtained by synthesizing a plurality of motion compensation predicted image signals corresponding to bidirectional prediction of a B picture can be used as a predicted image signal. it can. As a result, the prediction error is reduced and the coding efficiency is greatly improved as compared with the case where synthesis is not used.

  In addition, in the conventional bi-prediction of a B picture, two motion vectors are required, and the information amount of the motion vector increases as compared with the unidirectional prediction. On the other hand, in a method of generating a plurality of motion compensated prediction image signals using three filters and selecting one prediction image signal from among them, such as the moving image decoding apparatus 300, the motion vector is 1 Since the book is sufficient, the information amount of the motion vector does not increase. Therefore, the problem that the prediction error becomes small but the code amount of the motion vector increases more than the conventional B picture bi-directional prediction is solved, and the prediction image signal with high accuracy without increasing the motion vector information is solved. Can be generated.

  Note that the present invention is not limited to the above-described embodiment. For example, the number of motion compensated prediction images as a generation element of a composite image is not limited to two, and may be three or more. It is also apparent that the present invention can be applied to a mode in which a motion vector is automatically determined by a predetermined calculation method and motion vector information is not transmitted, such as the MPEG-4 AVC direct mode and skip mode. is there. For example, with respect to a motion vector automatically determined by a predetermined calculation method, a predicted image is generated by transmitting only image synthesis mode information, and not only a motion vector but also an image synthesis mode is automatically determined. For example, a predicted image may be generated.

  The present invention can also be applied to B pictures. For example, for MPEG-4AVC B pictures, reference images can be selected from two lists (L0, L1), and the respective reference images can be combined. By applying the present invention, it is possible to synthesize images for each list (L0, L1) and further synthesize an image with the synthesized images. That is, it is possible to combine the images selected in each list, such as MPEG-4AVC B pictures, after combining the images in each list according to the present invention.

  As described above, the present invention is characterized in that a plurality of motion compensated prediction images are generated using one motion vector, and the images are combined to form a prediction image. Are all included in the present invention.

  The present invention also includes a moving picture coding program for causing a computer to realize the functions of the moving picture coding apparatuses 100 and 200 described above. This moving image encoding program may be read from a recording medium and taken into a computer, or may be transmitted via a communication network and taken into a computer. Furthermore, the present invention can also be applied to a transmission apparatus that packetizes and transmits the bit stream generated by the moving image encoding apparatuses 100 and 200 described above.

  The present invention relates to a moving image encoding apparatus and a moving image encoding program, and is particularly useful as a technique for improving the encoding efficiency of moving image encoding.

100, 200 Video encoding device 101 Image signal input terminals 102, 201 Local decoding unit 103 Subtractor 104 Orthogonal transformation / quantization unit 105 Inverse quantization / inverse orthogonal transformation unit 106 Adder 107 Reference image memory 108, 109, 110 Upsampling filter 111 Motion compensation / image synthesis selection unit 112 Motion vector detection unit 113 Entropy encoding unit 114 Bit stream output terminal 202 Motion compensation unit 203, 204 Filter 205 Image synthesis / selection unit

Claims (6)

A reference image signal is generated by locally decoding an encoded signal obtained by encoding an encoding target image signal in an input moving image signal, and a prediction image signal obtained by performing motion compensation prediction on the reference image signal is used. In the moving picture coding method for performing motion compensation predictive coding of the current coding target picture signal in the input moving picture signal,
A first step of detecting a motion vector of a current encoding target image signal in the input moving image signal;
A second step of generating a motion compensated prediction image signal corresponding to each of the filter processes from the generated reference image signal by two or more filter processes and a motion compensation process based on the same motion vector; When,
A third step of generating one or more combined motion compensated prediction image signals by combining two or more motion compensation prediction image signals of different combinations among the plurality of generated motion compensation prediction image signals;
A fourth step of selecting a signal having the smallest error evaluation value from the plurality of motion compensated prediction image signals and the combined motion compensation prediction image signal and outputting the selected signal as the prediction image signal;
A differential signal between the current encoding target image signal and the predicted image signal selected in the fourth step is orthogonally transformed and quantized to generate a quantized signal that is the encoded signal. And the steps
Reference image signal specifying information, motion vector specifying information, and filter for specifying each of the generated reference image signal, the same motion vector, and the filter processing used for generating the selected predicted image signal A sixth step of adding processing specific information to a moving image encoded signal generated by performing a predetermined encoding process on the quantized signal;
A seventh step of generating a new reference image signal by local decoding processing by adding the selected predicted image signal to a signal obtained by inverse quantization and inverse orthogonal transform of the quantized signal; A moving picture encoding method comprising: In the third step, one or more synthesized motion compensated predictions are obtained by weighting and synthesizing two or more motion compensated predicted image signals of different combinations among the plurality of motion compensated predicted image signals at a set synthesis ratio. Generate an image signal,
The sixth step is a composite ratio information indicating a composite ratio used for generating the predicted image signal when the fourth step selects the composite motion compensated predicted image signal and outputs it as the predicted image signal. The video encoding method according to claim 1, further comprising: adding to the video encoded signal. A reference image signal is generated by locally decoding an encoded signal obtained by encoding an encoding target image signal in an input moving image signal, and a prediction image signal obtained by performing motion compensation prediction on the reference image signal is used. In the moving picture coding apparatus for performing motion compensation predictive coding of the current coding target picture signal in the input moving picture signal,
Motion vector detecting means for detecting a motion vector of the current encoding target image signal in the input moving image signal;
A motion-compensated prediction image that generates a motion-compensated prediction image signal corresponding to each filter process from the generated reference image signal by two or more filter processes and a motion compensation process based on the same motion vector Signal generating means;
A combined motion compensated prediction image signal generation unit that generates two or more combined motion compensation prediction image signals by combining two or more motion compensation prediction image signals of different combinations among the plurality of generated motion compensation prediction image signals; ,
Prediction image signal selection means for selecting a signal having the smallest error evaluation value from the plurality of motion compensation prediction image signals and the combined motion compensation prediction image signal and outputting the selected signal as the prediction image signal;
Quantization that generates a post-quantization signal that is the encoded signal by performing orthogonal transform and quantization on a difference signal between the current encoding target image signal and the predicted image signal selected by the predicted image signal selection unit A signal generation means after conversion;
Reference image signal specifying information, motion vector specifying information, and filter for specifying each of the generated reference image signal, the same motion vector, and the filter processing used for generating the selected predicted image signal Encoding means for adding processing specific information to a moving image encoded signal generated by performing a predetermined encoding process on the quantized signal; and
Reference image signal generation means for generating a new reference image signal by local decoding processing by adding the selected predicted image signal to a signal obtained by inverse quantization and inverse orthogonal transform of the quantized signal A moving picture encoding apparatus comprising: The synthesized motion compensated predicted image signal generation means weights and synthesizes each of two or more motion compensated predicted image signals of different combinations among the plurality of motion compensated predicted image signals at a set synthesis ratio. Generating a composite motion compensated prediction image signal;
The encoding means, when the prediction image signal selection means selects the composite motion compensated prediction image signal and outputs it as the prediction image signal, the composition ratio information indicating the composition ratio used to generate the prediction image signal The video encoding apparatus according to claim 3, further comprising: a video encoding signal added to the video encoding signal. A reference image signal is generated by locally decoding an encoded signal obtained by encoding an encoding target image signal in an input moving image signal, and a prediction image signal obtained by performing motion compensation prediction on the reference image signal is used. In the moving picture coding program for causing the computer to execute a moving picture coding process for performing motion compensation prediction coding of the current coding target image signal in the input moving picture signal,
In the computer,
A first step of detecting a motion vector of a current encoding target image signal in the input moving image signal;
A second step of generating a motion compensated prediction image signal corresponding to each of the filter processes from the generated reference image signal by two or more filter processes and a motion compensation process based on the same motion vector; When,
A third step of generating one or more combined motion compensated prediction image signals by combining two or more motion compensation prediction image signals of different combinations among the plurality of generated motion compensation prediction image signals;
A fourth step of selecting a signal having the smallest error evaluation value from the plurality of motion compensated prediction image signals and the combined motion compensation prediction image signal and outputting the selected signal as the prediction image signal;
A differential signal between the current encoding target image signal and the predicted image signal selected in the fourth step is orthogonally transformed and quantized to generate a quantized signal that is the encoded signal. And the steps
Reference image signal specifying information, motion vector specifying information, and filter for specifying each of the generated reference image signal, the same motion vector, and the filter processing used for generating the selected predicted image signal A sixth step of adding processing specific information to a moving image encoded signal generated by performing a predetermined encoding process on the quantized signal;
A seventh step of generating a new reference image signal by local decoding processing by adding the selected predicted image signal to a signal obtained by inverse quantization and inverse orthogonal transform of the quantized signal; A moving picture encoding program characterized in that In the third step, one or more synthesized motion compensated predictions are obtained by weighting and synthesizing two or more motion compensated predicted image signals of different combinations among the plurality of motion compensated predicted image signals at a set synthesis ratio. Generate an image signal,
The sixth step is a composite ratio information indicating a composite ratio used for generating the predicted image signal when the fourth step selects the composite motion compensated predicted image signal and outputs it as the predicted image signal. The moving picture coding program according to claim 5, further adding to the moving picture coded signal.

Images