JP2016032182A - Dynamic image encoder, dynamic image decoder, dynamic image encoding/decoding method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a dynamic image encoder capable of achieving high encoding efficiency, allowing intra-prediction against an inter-prediction residual signal.SOLUTION: A secondary intra-prediction part 104 predicts against an inter-prediction residual signal in a predetermined prediction direction per block. A secondary non-casual prediction part 105 predicts each signal belonging to each block against the inter-prediction residual signal on the basis of peripheral signals. A selection part 106 determines with which of the secondary intra-prediction part 104 and the secondary non-casual prediction part 105 a prediction is performed per block.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a moving image encoding device, a moving image decoding device, a moving image encoding / decoding method, and a program, and in particular, a moving image encoding device and a moving image decoding that allow intra prediction for an inter prediction residual signal. The present invention relates to an apparatus, a moving image encoding / decoding method, and a program.

  Non-Patent Document 1 discloses a standard for video compression coding. In the same standard, when performing inter-frame prediction (hereinafter referred to as inter prediction), orthogonal transform and encoding are performed on a motion compensation residual signal.

  On the other hand, Non-Patent Document 2 discloses a method of performing intra-frame prediction (hereinafter referred to as intra prediction) on a motion compensated residual signal for the purpose of improving coding performance by improving prediction performance. Since this method performs double prediction by further applying intra prediction to inter prediction, the same literature refers to the prediction as second order prediction.

"Draft ITU-T recommendation and final draft international standard of joint video specification (ITU-T Rec. H.264 / ISO / IEC 14 496-10 AVC)," in Joint Video Team (JVT) of ISO / IEC MPEG and ITU -T VCEG, JVTG050, 2003. S. Li, S. Chen, J. Wang, and L. Yu, "Second order prediction on H.264 / AVC," in PCS: 2009 Picture Coding Symposium, 2009, pp. 85-88, Picture Coding Symposium 2009, Chicago, IL, May 06-08, 2009. A. Asif and J. M. F. Moura, "Image Codec with Non-causal Prediction, Residual Mean Removal, and Cascaded VQ," IEEE Trans. On Video Tech., Vol. 6, no. 1, Feb. 1996.

  However, the prior art as described above has a problem in that there is still room for further improving the encoding efficiency.

  In inter prediction disclosed in Non-Patent Document 1 that discloses a standard, temporal correlation in a video can be reduced. However, since prediction in the spatial direction is not performed, redundancy of pixel values remains in the spatial direction. ing.

  In the method of Non-Patent Document 2, since the spatial prediction is performed on the inter prediction residual signal, the redundancy in the spatial direction can be reduced. However, in this method, in order to perform intra prediction of the standard adopted in Non-Patent Document 1, etc., it is necessary to generate intra prediction reference pixels outside the inter prediction block.

  Moreover, although the inter prediction residual signal remains redundant in the spatial direction, the correlation between pixel values between adjacent pixels is lower than that of a general video signal. In general, assuming that the correlation between pixel values between adjacent pixels is ρ (ρ is 1 or less), the correlation between pixels separated by N pixels is proportional to ρ to the Nth power. Therefore, in the method of Non-Patent Document 2 in which the reference pixel value as in Non-Patent Document 1 is applied to the inter prediction residual signal as a predicted value along the prediction direction, the prediction is performed as the distance from the reference pixel increases. Since the decrease in performance becomes significant, the effect of improving the encoding performance by applying intra prediction is limited.

  That is, Non-Patent Document 2 is a secondary prediction framework in which intra prediction is further applied to inter prediction. Although the encoding efficiency is improved as compared with the framework of Non-Patent Document 1 or the like, there is room for further improvement. Remaining. In particular, as a major problem of Non-Patent Document 2, an intra prediction method (method of Non-Patent Document 1) in which a reference pixel value is set as a predicted value along the prediction direction without considering the characteristics of the inter prediction residual signal. The point used is mentioned.

  Here, the intra prediction of Non-Patent Document 1 corresponds to a causal prediction framework in that an encoding target pixel (future pixel) is predicted from an encoded pixel (past pixel). On the other hand, there is a non-causal prediction framework disclosed in Non-Patent Document 3 for causal prediction.

FIG. 3 is a diagram conceptually showing causal prediction and non-causal prediction. In the causal prediction shown in [1], in general, the pixel x _{i, j at} the position (i, j) to be predicted is, for example, the pixel x _{i−1, j−1} , x _{i, j− 1} , x _{i-1, j} etc. As described above, since the prediction direction is limited, as the pixel positions are separated from each other as described above, the correlation is lowered, thereby reducing the encoding efficiency. On the other hand, in the non-causal prediction shown in [2], the prediction direction is not limited, and in general, the pixel x _{i, j at} the position (i, j) to be predicted is all the surrounding pixels, for example, the pixel x It is predicted from _{i, j-1} , x _{i-1, j} , x _{i + 1, j} , x _{i, j + 1 and the} like. In this way, compared to causal prediction, non-causal prediction has more pixels to be referenced for prediction and can use adjacent pixels in all directions. It is expected that the impact will be reduced and an excellent prediction can be made.

  However, the non-causal prediction disclosed in Non-Patent Document 3 is originally assumed to be used in a quadratic prediction framework that applies intra prediction to an inter prediction residual signal as in Non-Patent Document 2. Absent. In particular, the pixel is not an inter prediction residual signal, and it is necessary to process the entire image at once in order to improve the coding efficiency, which causes a problem that a large calculation cost is required. Specifically, it is as follows.

  That is, in Non-Patent Document 3, since the processing target signal is a pixel value, the rightward and downward pixels are unknown in the decoding device. Further, although the power of the prediction error signal of the non-causal method is small, the prediction function used at the time of decoding is non-orthogonal, so that the error may increase. Therefore, it is known that the calculation load becomes heavier than a decoding device based on causal prediction in order to suppress the error expansion. When decoding is actually performed, an iterative processing algorithm or a recursive algorithm must be used to maintain coding efficiency.

  Non-Patent Document 3 uses a method called a recursive framework. The encoding device and the decoding device each perform a recursive operation to realize high-efficiency encoding. Specifically, the following steps 1 to 4 are performed on the encoding side.

[Step 1] The following four types of parameters are obtained for the entire image as a processing target, and an optimal parameter set is repeatedly searched for each row of the image.
(1) Average value of all screens, (2) Zero mean image, (3) Power of whole image (sample power), (4) Correlation coefficient (sample correlations) of image.
[Step 2] Generate a recursive matrix using the parameter set obtained in Step 1.
[Step 3] A prediction error image is generated from the last row using a recursive matrix, and the prediction error image is quantized. Thus, what is sent to the decoding device is a value after entropy coding of the quantized prediction error image and a parameter of non-causal prediction (obtained in step 1).

The decoding apparatus performs the following steps 1 to 3 in accordance with the above processing of the encoding apparatus.
[Step 1] The obtained signal is entropy decoded and dequantized to generate an error image.
[Step 2] Based on the obtained parameter information, the same recursive matrix as the output of step 2 on the encoding side is calculated.
[Step 3] A reconstruction matrix is generated by a recursive matrix, and one row is reconstructed at a time from the last row of the image.

  As described above, the non-causal model in-screen prediction in Non-Patent Document 3 uses recursive processing for the entire screen in order to maintain encoding efficiency (to optimize parameters for the entire screen). Since the processing must be performed collectively, the computational complexity becomes very high at the time of implementation. In particular, real-time processing is considered impossible for ultra-high resolution images.

  In view of the above-described problems of the prior art, an object of the present invention is to provide a moving picture coding apparatus that can achieve high coding efficiency and allows intra prediction for an inter prediction residual signal. .

  Another object of the present invention is to provide a moving picture decoding apparatus corresponding to the moving picture encoding apparatus.

  Furthermore, an object of the present invention is to provide a moving image encoding / decoding method and program that allow intra prediction on an inter prediction residual signal, which can achieve high encoding efficiency.

In order to achieve the above object, the present invention is a video encoding apparatus that allows intra prediction for an inter prediction residual signal, and for each block in a predetermined prediction direction with respect to the inter prediction residual signal. A secondary intra-prediction unit that performs prediction, a secondary non-causal prediction unit that predicts each signal belonging to each block based on surrounding signals, with respect to the inter prediction residual signal,
A selection unit that determines whether to perform prediction by the secondary intra prediction unit or the secondary non-causal prediction unit for each block.

  The present invention is also a moving picture decoding apparatus that allows intra prediction on an inter prediction residual signal, and that performs prediction in a predetermined prediction direction for each block on the inter prediction residual signal. A prediction unit, a secondary non-causal prediction unit that predicts each signal belonging to each block based on surrounding signals with respect to the inter prediction residual signal, and the secondary intra prediction unit for each block Or a selection unit that determines which of the secondary non-causal prediction units performs the prediction.

  In addition, the present invention is a moving picture encoding / decoding method that allows intra prediction for an inter prediction residual signal, and has a predetermined prediction direction for each block with respect to an inter prediction residual signal to be encoded. The encoding-side secondary intra prediction stage that performs prediction at the encoding side, and the encoding-side secondary prediction that predicts each signal belonging to each block based on the surrounding signals with respect to the inter prediction residual signal to be encoded A non-causal prediction stage and an encoding side selection for selecting whether to perform prediction by the encoding side secondary intra prediction stage or the encoding side secondary non-causal prediction stage for each block to be encoded A step of encoding each block to be encoded by applying the selected prediction to form a bit stream, a decoding step of decoding the bit stream, and an interface to be decoded A decoding-side secondary intra prediction stage that performs prediction in a predetermined prediction direction for each block with respect to the prediction residual signal, and each signal belonging to each block with respect to the inter prediction residual signal to be decoded Prediction on the decoding side secondary non-causal prediction stage for prediction based on surrounding signals, and for each block to be decoded, prediction by either the decoding target secondary intra prediction unit stage or the decoding target secondary non-causal prediction stage And a decoding-side selection step of selecting whether to perform.

  Further, the present invention is a program for causing a computer to execute the moving image encoding / decoding method.

  According to the present invention, when applying intra prediction to an inter prediction residual signal, either a secondary intra prediction unit or a secondary non-causal prediction unit can be selected. By applying the prediction means, high encoding efficiency can be achieved.

It is a functional block diagram of the moving image encoder which concerns on one Embodiment. It is a functional block diagram of the moving image decoding apparatus which concerns on one Embodiment. It is a figure which shows a causal prediction and a noncausal prediction notionally. It is a figure for demonstrating the calculation of noncausal interpolation prediction. It is a figure for demonstrating the formula of a general noncausal prediction.

  FIG. 1 is a functional block diagram of a video encoding apparatus according to an embodiment. The video encoding apparatus 100 includes an intra prediction unit 101, an inter prediction unit 102, a motion compensation unit 103, a secondary intra prediction unit 104 (hereinafter referred to as an SOP unit 104 [Second Order Prediction]), a secondary non-causal prediction. Unit 105 (hereinafter referred to as NCSOP unit 105 [Non Causal Second Order Prediction]), selection unit 106, transform / quantization unit 107, inverse quantization / inverse transform unit 108, switch 109, adder 110, filter unit 111, A first memory 112, an encoding unit 113, a differentiator 120, and a differentiator 130 are provided.

  The video encoding apparatus 100 includes the configuration, reads the video to be encoded for each frame, divides the frame into blocks, and applies various predictions to the configured pixels for each block. Then, the residual is transformed and quantized and encoded to form a bit stream.

  FIG. 2 is a functional block diagram of a video decoding apparatus 200 according to an embodiment that decodes the bitstream encoded by the video encoding apparatus 100 of FIG. The moving picture decoding apparatus 200 includes an intra prediction unit 201, a motion compensation unit 203, a secondary intra prediction unit 204 (hereinafter referred to as SOP unit 204 [Second Order Prediction]), a secondary non-causal prediction unit 205 (hereinafter referred to as NCSOP). Unit 205 [Non Causal Second Order Prediction]), selection unit 206, inverse quantization / inverse transformation unit 208, switch 219, adder 210, filter unit 211, first memory 212, decoding unit 213, second memory 214 And a switch 229.

  Here, before explaining the details of each part of the moving picture coding apparatus 100 and the moving picture decoding apparatus 200, the overall features will be explained first.

  In each part of the moving picture coding apparatus 100 and the moving picture decoding apparatus 200, a function part name that is common to each other and a function part having a common lower two-digit reference number, for example, the intra prediction part 101 and the intra prediction part 201 The processing contents are the same.

  The characteristic configuration of the moving picture coding apparatus 100 is an SOP unit 104, an NCSOP unit 105, and a selection unit 106 (and a differentiator 130) shown as a functional block group B100. If all of these characteristic configurations are omitted, the moving image encoding apparatus 100 has the same configuration as that of the standard disclosed in Non-Patent Document 1. Also, if only the NCSOP unit 105 is omitted among these characteristic configurations, the moving picture coding apparatus 100 has the same configuration as that disclosed in Non-Patent Document 2.

  With this characteristic configuration, the video encoding apparatus 100 performs intra-screen prediction based on non-causal interpolation prediction proposed in Non-Patent Document 3 as secondary prediction for the inter prediction residual signal. The intra coding method used for secondary prediction is adaptively selected according to the characteristics of the inter prediction residual signal. Thus, further improvement of the prediction performance of the secondary prediction method (method of Non-Patent Document 2) that enables reduction of redundancy in the spatial direction in the inter prediction residual is achieved.

  That is, in the method of Non-Patent Document 2, since only the SOP unit 104 is used, the encoding efficiency may be insufficient depending on the characteristics of the inter prediction residual signal. The encoding efficiency is improved by selectively using the NCSOP unit 105 suitable for the signal characteristics.

  Similarly, the characteristic configuration of the moving picture decoding apparatus 200 is an SOP unit 204, an NCSOP unit 205, and a selection unit 206 shown as a functional block group B200. If all of these characteristic configurations are omitted, the moving picture decoding apparatus 200 has the same configuration as the standard disclosed in Non-Patent Document 1. Further, if only the NCSOP unit 205 is omitted among these characteristic configurations, the moving picture decoding apparatus 200 has the same configuration as that disclosed in Non-Patent Document 2.

  With the characteristic configuration, the moving picture decoding apparatus 200 achieves further improvement in prediction performance of the secondary prediction method (method of Non-Patent Document 2) that enables reduction of redundancy in the spatial direction in the inter prediction residual. Decoding processing corresponding to the image encoding device 100 is possible.

  Hereinafter, the details of each part of FIGS. 1 and 2 will be described. Note that, as described above, many of the units of the video encoding device 100 and the video decoding device 200 have the same processing contents themselves, but have different data flows. Therefore, the functional unit having the common processing content will be described in parallel by the moving image encoding device 100 and the moving image decoding device 200, and different data flows will be described separately. For convenience of description in parallel, the moving image encoding device 100 is abbreviated as an encoder, and the moving image decoding device 200 is abbreviated as a decoder.

  The intra prediction units 101 and 201 perform intra-frame prediction processing in Non-Patent Document 1. In the case of intra prediction encoding (also referred to as intra prediction encoding), it is performed in units of square blocks. In the case of an N × N block, the right end N pixels in the left block, the lower end N pixels in the upper block, the lower end 4 pixels in the upper right block, and several pixels in the lower right one pixel in the upper left block are included in the block. N × N pixel values are predicted and encoded. There are several prediction modes depending on the location and usage of neighboring encoded pixels to be used. A prediction value based on a prediction direction in which the highest coding performance is expected is output.

  As a flow common to the encoder and decoder, the intra prediction units 101 and 201 refer to the first memories 112 and 212 that hold the encoded pixels (decoded pixels) of the prediction reference, and output the predicted values. In the encoder, the predicted value is output to the difference unit 120, and in the decoder, the predicted value is output to the adder 210 via the switches 219 and 229. The encoder further outputs a prediction mode (intra prediction information) to the encoding unit 113.

  The inter prediction unit 102 determines a motion vector used for interframe prediction. The same unit evaluates a block suitable for prediction in the reference frame with respect to the processing block, expresses a position optimal for prediction and a spatial displacement of the processing block as a vector, and uses it as a motion vector. Here, the first memory 112 holds the reference frame. As an output of the same unit, a motion vector is output and passed to the motion compensation unit 103 and the encoding unit 113.

  The motion compensation units 103 and 203 use the motion vector estimated as described above to generate a prediction value from the corresponding portion of the reference frame held by the first memories 112 and 212. The generated prediction value is output as the output of the same part, and is output to the differencer 120 in the encoder and to the adder 210 via the switches 219 and 229 in the decoder.

  On the decoder side, when a prediction signal is sent from either the intra prediction unit 201 or the motion compensation unit 203, the switch 219 sends the prediction signal to the switch 229. When the prediction signal is sent from either the switch 219 or the functional block group B200, the switch 229 sends the prediction signal to the adder 210.

  The SOP unit 104 performs secondary prediction (application of intra prediction to inter prediction residuals) proposed in Non-Patent Document 2. That is, intra-frame prediction is performed on the inter prediction residual (residual signal by motion compensation). Specifically, with respect to the residual signal for motion compensation, the difference between the decoded pixel values at the upper, left, upper left, and upper right of the processing block and the decoded pixel values at the upper, left, upper left, and upper right of the motion compensation reference block is calculated. Intraframe prediction is performed using the predicted value.

  Here, the inter prediction residual signal to be predicted is the inter prediction signal obtained by reading the encoding target block and processing it by the inter prediction unit 102 and the motion compensation unit 103, the signal of the encoding target block itself, Are generated by the differentiator 120.

  The inter prediction residual signal to be referred to is also output to the transform / quantization unit 107 in parallel with the output of the differencer 120 to the SOP unit 104, and the transform / quantization unit 107 Then, after passing through the inverse quantization / inverse transform unit 108, the flow is switched toward the line L100 by the switch 109, so that the first memory 112 holds it. This makes it possible to provide a reference inter prediction residual signal when the frame of the current block to be encoded becomes an inter prediction reference frame at a later time.

  As the output of the SOP unit 104, a value obtained by adding the prediction value obtained by motion compensation and the intra-frame prediction value is output as a prediction value, and the difference unit 130 obtains the difference between the prediction value and the signal to be encoded. To the transform / quantization unit 107. Note that the prediction value obtained by motion compensation may be obtained by receiving the result obtained by the motion compensation unit 103 before inputting the current encoding target block to the differencer 120 described above. (The flow of data exchange at the time of receipt is not shown in FIG. 1)

  On the other hand, the corresponding SOP unit 204 on the decoder side generates a prediction value on the decoding side based on the SOP shown in Non-Patent Document 2. In the same section, the difference between the decoded pixels on the upper, left, upper left, and upper right of the processing block and the upper, left, upper left, and upper right decoded pixels indicated by the motion vector is used as a reference pixel for intra prediction. The prediction value is generated according to the intra prediction direction of the control information. Next, a value obtained by adding the above-described predicted value and the predicted value by motion compensation is set as a predicted value of SOP. As the output of the same part, the predicted value of SOP is output and passed to the adder 210 via the switch 229.

  In order for the SOP unit 204 on the decoder side to perform the process, an inter prediction residual signal as a reference pixel is required, and this is stored in the first memory 212 by the motion compensation unit 203 by the same process as the encoder side. Keep it. For this reason, on the decoder side, the motion compensation unit 203 obtains a prediction signal and refers to the first memory 212 to obtain an inter prediction residual signal as a difference between the obtained prediction signal and the reference signal, which is shown as a line L200. As described above, it is assumed that the data is stored in the first memory 212 again.

  The difference value obtained by the differentiator 130 from the prediction value obtained by the SOP 104 on the encoder side is encoded by the transform / quantization unit 107 and the encoding unit 113, and is then transmitted to the decoder side to be decoded by the decoding unit 213 and the inverse quantum. The conversion unit 208 returns to the difference value, and the adder 210 adds the prediction value obtained by the decoder-side SOP unit 204 to obtain a decoded pixel value.

  The transform / quantization unit 107 receives the prediction error signal and performs orthogonal transform and quantization on the signal. As an output of the same unit, a quantized transform coefficient is output and passed to the inverse quantization / inverse transform unit 108 and the encoding unit 113. The prediction that can be applied to the input prediction error signal is performed by any of the intra prediction unit 101, the motion compensation unit 103, the SOP unit 104, or the NCSOP unit 105.

  In the inverse quantization / inverse transform units 108 and 208, the quantized transform coefficients are input. In the same part, inverse quantization and inverse transformation are performed on the input signal to the same part to obtain a residual signal. The obtained residual signal is output as an output of the same part, and is passed to the switch 109 on the encoder side and to the adder 210 on the decoder side. 2, on the decoder side, the residual signal is passed to the NCSOP unit 205 via the selection unit 206 only in the case of a residual to which the NCSOP unit 205 is applied, as indicated by a one-dot chain line L250 in FIG.

  As described in the SOP unit 104, the switch 109 passes an inter prediction residual signal to be used as a prediction reference in the SOP unit 104 to the first memory 112 as indicated by a line L100, and other types of remaining signals. For the difference signal, a switching process of passing to the adder 110 is performed.

  The adders 110 and 210 add the residual sent from the switch 109 or the inverse quantization / inverse transform unit 208 and the prediction signal sent from each prediction unit to obtain a reconstructed pixel signal (decoded image signal). To the filter units 111 and 211.

  The filter units 111 and 211 are filters for reducing block distortion that occurs during image encoding prior to storing decoded images in the first memories 112 and 212. The decoded image after the filter processing is output as the output of the same unit.

  The encoding unit 113 performs entropy encoding on the quantized coefficient, encoding mode, motion vector information, and the like. An encoded bit stream is output as the output of the same unit.

  The decoding unit 213 performs entropy decoding on the bit stream obtained by the encoding unit 113, and acquires control information such as an encoding mode and a residual signal. As the output of the same unit, the acquired control information and residual signal are output. The residual signal is passed to the inverse quantization / inverse transform unit 208, and the control information is passed to the second memory 214.

  The first memories 112 and 212 are memories for storing decoded images, and supply decoded pixel values to the intra prediction units 101 and 201, motion compensation units 103 and 203, and SOP units 104 and 204 as necessary.

  The second memory 214 is a memory for accumulating control information (prediction mode information, etc.), and is necessary when various predictions are performed by the intra prediction unit 201, the motion compensation unit 203, the SOP unit 204, and the like. In response, control information is supplied.

  The NCSOP units 105 and 205, which are characteristic configurations in the present invention and are selectively applied by the selection units 106 and 206, will be described below.

  First, the selection unit 106 performs intra prediction encoding on the motion compensation residual signal, and enables selection of an appropriate prediction method from a plurality of candidates for the intra prediction method. In particular, it is determined whether the SOP unit 104 or the NCSOP unit 105 should be adopted in the encoding processing unit.

  The determined information is sent to the encoding unit 113 as control information (although the flow is not shown in FIG. 1), is encoded, and is added to the bitstream. The information is decoded and received by the selection unit 206 on the decoder side, so that the decoder side similarly determines which of the SOP unit 204 and the NCSOP unit 205 is adopted.

  If one of the predictions is expected to be effective for the entire slice, identification information is given to the slice header, and only one prediction indicated by the identification information is used in the slice. Further, as a result of this evaluation, it can be shown by identification information that neither is used in the slice. Furthermore, it is also possible to identify using either one or neither using an encoding syntax higher than the slice header.

  If one of the slices is not determined, identification information indicating which prediction is used is given to each block in a block in which intra prediction is performed on a motion compensation residual signal.

  As a method of determining one of them in units of slices, a method using a variance value of pixel values in a slice is possible. As a method of deciding either one in units of blocks, it is possible to perform the encoding performance based on the prediction value of the variance value of the pixel value of the processing block, the variance value of the motion compensation residual signal and the correlation between pixels.

  As the criteria to be determined in each unit, the variance value of the pixel value of the processing block, the variance value of the motion compensation residual signal and the inter-pixel correlation, and the encoding performance based on the respective prediction (based on the encoding error and the generated code amount) Various encoding performance evaluation indexes such as encoding performance evaluated by the user can be used. The magnitude of the variance value or the correlation between pixels may be determined based on, for example, a threshold value. The encoding performance may be determined based on the actual evaluation and the higher performance.

  In consideration of the nature of intra prediction based on the prediction direction used in the SOP unit 104, there is a concern that the prediction performance may be significantly deteriorated with complex textures. Therefore, the NCSOP unit 105 is suitable in a region where the variance value is high or a region where the correlation between pixels is low.

  The NCSOP unit 105 generates a prediction value by using the non-causal interpolation prediction proposed in Non-Patent Document 3 as intra-screen prediction for the residual signal of motion compensation by a method specialized in the present invention. . Details of the predicted value generation method in the same part will be described later.

  In non-causal interpolation prediction in the NCSOP unit 105, matrix calculation is performed on the inter-frame prediction residual signal to be predicted using an interpolation prediction function. An encoding procedure will be described by taking an image configured as an N pixel × M line matrix in FIG. 4 as an example.

FIG. 4 shows an example of non-causal interpolation prediction processing in an N pixel × M line coding block. In the figure, using the prediction error signal values x ₂ , x _{N + 1} , x _{N + 3} , x _{2N + 2} at pixel coordinates 2, N + 1, N + 3, 2N + 2, pixel coordinates N + 2 An example of calculating the encoding target value y _{N + 2} in is shown.

In FIG. 4, the pixels at the four gray corners are pixels for which the prediction process is not performed. In this pixel, the residual signal of the motion compensation prediction (which is the primary prediction) is entropy encoded. White pixels are pixels for which prediction processing is performed. In the non-causal interpolation prediction of the second-order prediction, the residual signal value X = {x _n | n = 1, 2,... NM} of motion compensated prediction at each pixel in the block is used. The encoding target value Y = {y _n | n = 1, 2,..., NM} is calculated, and entropy encoding is performed on Y.

For example, for pixels N + 2, 2N, (M-1) (N + 3), using residual signal values xN _{+ 2} , _x2N , x _{(M-1) N + 3} , encoding target values The process of calculating y _{N + 2} , y _2N , y _{(M-1) N + 3} is

It becomes. When all the pixels are predicted in this way, a general non-causal prediction can be expressed by Equation 4.

However, β _v , β _h , β _ld and β _rd are prediction coefficients for vertical, horizontal, left diagonal and right diagonal, respectively, and the value ranges within [0, 1] (0 to 1) Each predetermined value is used. Moreover, y _ij is a prediction error. Note that the notation of the coordinate position in Equation 4 is as shown in FIG. 5 (unlike FIG. 4).

  As described above, the non-causal prediction can be defined in advance as a predetermined linear sum using all or a part of the surrounding eight pixels of the target pixel. Furthermore, a non-causal prediction may be similarly defined using a predetermined range wider than that shown in FIG.

  Regarding the non-causal interpolation prediction, the encoding process and the decoding process for the above prediction process are shown in Expression 5 and Expression 6. That is, the NCSOP unit 106 on the encoder side performs the process of Equation 5, and the NCSOP unit 206 on the decoder side performs the process of Equation 6.

  Equation 5 shows predictive coding for a coded block represented by an N × M matrix L. That is, Expression 5 shows the processing of each pixel described in the above (1) to (3) and (4) as the calculation of the entire block.

  However, C is an NM × NM encoding matrix, vec (X) is a vector of all pixels in the NM dimension, and vec (E) is a vector after encoding in the NM dimension.

Equation 6 represents the decoding process. C ⁻¹ represents the inverse matrix of C described above.

  As is clear from the above description, the present invention proposes a non-causal interpolation intra prediction method for a plurality of block sizes in an SOP for removing spatial correlation with respect to an interframe residual signal. Compared to causal prediction, non-causal prediction is known to be able to perform superior prediction because there are more pixels to be referred to for prediction and adjacent pixels in all directions are used.

  In particular, the non-causal interpolation intra prediction method of the present invention differs from that of Non-Patent Document 3 in that only information on the matrix C and its inverse matrix is used on the encoder side and the decoder side as information fixed in advance. Thus, it is only necessary to perform the operation in units of blocks to be encoded / decoded. Therefore, iterative calculation or the like is unnecessary, and encoding and decoding can be performed with a low calculation load.

  Based on the above, the decoder side can operate as follows.

  The selection unit 206 enables intra prediction for the inter prediction residual signal, and selects one prediction method from a plurality of intra prediction methods according to the prediction information described in the bitstream. That is, use of either the SOP unit 104 or the NCSOP unit 105 is selected.

  As described above, the NCSOP unit 205 generates a predicted value based on the non-causal interpolation intra prediction method unique to the present invention. That is, in the same section, with respect to the residual signal output from the decoding section 213 and obtained by being processed by the inverse quantization / inverse transform section 208 (in FIG. 2, the flow is indicated by a one-dot chain line). Then, the decoding process of non-causal interpolation prediction is performed by multiplying the inverse matrix of the encoding matrix C of Equation 6. By adding a motion compensation prediction value to the addition unit 210 in the result, a decoded value of the processing block can be obtained.

  (Supplementary Item 1) The present invention can also be provided as a program that causes a computer to function as the moving image encoding device 100 or the moving image decoding device 200. The computer can employ a known hardware configuration such as a CPU (Central Processing Unit), memory, and various I / Fs, and the CPU executes instructions corresponding to the functions of the respective units shown in FIG. 1 or FIG. Will be.

  (Supplementary Item 2) The present invention can also be provided as a moving image encoding / decoding system including the moving image encoding device 100 and the moving image decoding device 200. In this case, in the system, the bit stream encoded by the encoding unit 113 by the moving image encoding device 100 is received by the decoding unit 213 of the moving image decoding device 200 and decoded. The operations of the video encoding device 100 and the video decoding device 200 in the system are the same as described above.

  (Supplementary Item 3) The present invention can also be provided as an operation method (moving image encoding / decoding method) of a moving image encoding / decoding system including the moving image encoding device 100 and the moving image decoding device 200 described above. The moving image encoding / decoding method can be provided as a program for causing a computer to execute the method. The hardware configuration of the computer is the same as that described in (Supplementary Item 1).

  DESCRIPTION OF SYMBOLS 100 ... Video encoding apparatus, 200 ... Video decoding apparatus, 101, 201 ... Intra prediction part, 102, 202 ... Inter prediction part, 103, 203 ... Motion compensation part, 104, 204 ... Secondary intra prediction part, 105, 205 ... Secondary non-causal prediction part , 106,206 ... selection unit, 107 ... transformation / quantization unit, 108,208 ... inverse quantization / inverse transformation unit, 110,210 ... adder, 111,211 ... filter unit, 112,212 ... first memory, 214 ... second memory, 113 ... encoding 213 ... decoding unit 120,130 ... differentiator 109,219,229 ... switch

Claims (12)

A video encoding device that allows intra prediction for an inter prediction residual signal,
For the inter prediction residual signal, a secondary intra prediction unit that performs prediction in a predetermined prediction direction for each block;
For the inter prediction residual signal, a secondary non-causal prediction unit that predicts each signal belonging to each block based on its surrounding signals,
A moving image encoding apparatus comprising: a selection unit that selects whether to perform prediction by the secondary intra prediction unit or the secondary non-causal prediction unit for each block.   The moving image encoding apparatus according to claim 1, wherein the secondary non-causal prediction unit predicts each signal belonging to each block based on a linear sum of surrounding signals.   The said selection part evaluates each encoding performance expectation of the said secondary intra prediction part or the said secondary noncausal prediction part, The said selection is based on the said evaluation result, The selection is characterized by the above-mentioned. The moving image encoding apparatus described in 1.   The selection unit uses the encoding performance evaluation based on the encoding error and the generated code amount, the dispersion value of the pixel value in the processing unit, or the inter-pixel correlation in the processing unit when evaluating the prediction of the encoding performance. The moving picture encoding apparatus according to claim 3, wherein   2. The information processing apparatus according to claim 1, wherein which of the secondary intra prediction unit and the secondary non-causal prediction unit is selected by the selection unit is encoded as identification information for each processing unit. 5. The moving image encoding apparatus according to any one of 4 to 4.   6. The moving picture coding apparatus according to claim 5, wherein the processing unit is given as a combination of one or more of a block, a slice, and a higher-level coding control syntax. A video decoding device that allows intra prediction for an inter prediction residual signal,
For the inter prediction residual signal, a secondary intra prediction unit that performs prediction in a predetermined prediction direction for each block;
For the inter prediction residual signal, a secondary non-causal prediction unit that predicts each signal belonging to each block based on its surrounding signals,
A moving image decoding apparatus comprising: a selection unit that selects whether to perform prediction by the secondary intra prediction unit or the secondary non-causal prediction unit for each block.   The moving image decoding apparatus according to claim 7, wherein the secondary non-causal prediction unit predicts each signal belonging to each block based on a linear sum of surrounding signals. Which of the secondary intra prediction unit or the secondary non-causal prediction unit should be selected by the selection unit is encoded as identification information for each processing unit,
The video decoding device according to claim 7 or 8, wherein the selection unit selects either the secondary intra prediction unit or the secondary non-causal prediction unit according to the identification information.   The moving picture decoding apparatus according to claim 9, wherein the processing unit is given as a combination of one or more of a block, a slice, and a higher-level encoding control syntax. A video encoding / decoding method that allows intra prediction for an inter prediction residual signal,
A coding-side secondary intra prediction stage that performs prediction in a predetermined prediction direction for each block on the inter prediction residual signal to be encoded;
For the inter-prediction residual signal to be encoded, the encoding-side secondary non-causal prediction stage that predicts each signal belonging to each block based on its surrounding signals,
For each block to be encoded, an encoding side selection step for selecting whether to perform prediction by the encoding side secondary intra prediction step or the encoding side secondary non-causal prediction step;
Applying the selected prediction to perform encoding for each block to be encoded into a bitstream; and
A decoding step of decoding the bitstream;
A decoding-side secondary intra prediction stage that performs prediction in a predetermined prediction direction for each block on the inter prediction residual signal to be decoded;
A decoding-side secondary non-causal prediction stage for predicting each signal belonging to each block based on the surrounding signals for the inter prediction residual signal to be decoded;
For each block to be decoded, a decoding side selection step of selecting whether to perform prediction by the decoding target secondary intra prediction unit step or the decoding target secondary non-causal prediction step is provided. Video encoding / decoding method.   A program for causing a computer to execute the moving image encoding / decoding method according to claim 11.