JP2008054182A - Dynamic image predictive coding method and apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a dynamic image predictive coding method and apparatus capable of performing coding improved in coding efficiency or image quality by applying an intra-frame predictive coding system. <P>SOLUTION: The disclosed method includes the steps of: reading pixel blocks from a dynamic image in a predetermined order (S1); coding the read pixel blocks in a plurality of prediction modes of different reference pixels (S2, S3); and calculating coding costs for each of the pixel blocks for each of the prediction modes (S4). The coding step includes: performing predictive coding in the plurality of prediction modes while being reflected with a prediction result for each combination of prediction modes selectable for each forward pixel block that affects the coding costs of a pixel block to be coded; and further, determining the prediction mode for each forward pixel block based on the combination of prediction modes minimizing the total sum of coding costs (9). <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a moving picture predictive coding method and apparatus, and more particularly, to a moving picture predictive coding method and apparatus suitable for an intraframe predictive coding method (intra prediction).

  H.264 / AVC is attracting attention as an international standard for moving picture coding. In H.264 / AVC, similar to the existing MPEG-2 / 4, intra-frame prediction coding, inter-frame prediction coding, discrete cosine transform and entropy coding are combined, but for intra-frame prediction coding, Compared to the conventional MPEG-2 / 4 prediction in the frequency domain (orthogonal transform coefficient), H.264 reduces the code amount by performing prediction in the spatial domain. .

  In H.264 / AVC intra-frame prediction, two types of prediction methods (16 × 16 prediction and 4 × 4 prediction) are defined for luminance signals, and either one is selected in units of macroblocks. One type of prediction method is defined for the color difference signal. In any prediction method, multiple prediction modes with different reference pixels (prediction directions) are provided.

  (1) 16 × 16 prediction of a luminance signal is a method for predicting a 16 × 16 pixel block (macroblock) at a time. As shown in FIG. 8, mode 0 for predicting vertically, predicting horizontally Mode 1 in which prediction is performed using an average value of a total of 32 pixels of the upper block and the left block as a predicted value, and mode in which the pixel values of the upper block and the left block are interpolated in a diagonal direction to obtain a predicted value In total, four prediction modes are prepared, and one of the prediction modes is selected in units of macroblocks. Note that only the pixels located within the screen are used for the macroblocks at the top and left ends of the screen, and prediction from outside the screen is not performed.

  (2) The 4 × 4 prediction of the luminance signal is a method of predicting by dividing a 16 × 16 pixel block in a macro block into 16 blocks each composed of 4 × 4 pixel blocks. One of the nine prediction modes shown in FIG. 9 is selected in units of blocks. Pixels that are at the top or left end of the screen or have not yet been decoded cannot be used for prediction.

  (3) The color difference signal is predicted by selecting one of the four prediction modes shown in FIG. 10 for 8 pixels × 8 lines in the macroblock.

Non-Patent Document 1 proposes an image coding method that improves coding efficiency in consideration of not only a prediction error due to a coding mode but also a generated code amount. Patent Document 1 proposes an image coding method that improves coding efficiency by adding a term that takes into consideration visual characteristics and the like to the technique of Non-Patent Document 1 described above. Patent Document 2 discloses a technique for improving the coding efficiency by estimating a prediction error and a generated code amount from a variance and a quantization parameter of a prediction residual signal and using a set of estimation parameters suitable for a coding target. ing. Patent Document 3 discloses a technique for determining an encoding mode of each macroblock by solving a set of inequalities including prediction errors and generated code amounts by linear programming for all macroblocks in a frame. .
JP 2006-5466 A JP 2005-86249 A JP 2005-260576 A Gary J. Sullivan and Thomas Wiegand, “Rate-Distortion Optimization for Video Compression”, IEEE Signal Processing Magazine, Vol.15, No.6, pp.74-90, Nov. 1998.

  In a general image encoding method, in order to increase encoding efficiency, an encoding target is predicted by using temporal or spatial correlation. Since the coding mode strongly depends on the prediction reference source, it affects not only the coding efficiency of a single coding unit, but also the prediction accuracy in other coding units that refer to the coding unit. However, in Non-Patent Document 1, Patent Document 2, and Patent Document 3, an encoding mode is determined for each single encoding unit, and the influence on subsequent encoding units has not been considered. For this reason, when viewed from the whole image, there is a risk of falling into a local optimum solution.

  For example, in H.264 / AVC, a block encoded by an encoder is decoded inside the encoder and used for prediction of other blocks. Therefore, it affects the coding efficiency of other blocks. In particular, in a prediction method that predicts from surrounding decoded pixels using spatial correlation, if an encoding mode is selected by focusing only on the encoding efficiency in a single block, the encoding efficiency of adjacent blocks There are problems that adversely affect

  Since Patent Document 3 aims to optimize code amount allocation, the selection of a code mode is not the main purpose. In addition, since the encoding distortion function having the same inclination is assumed for the same quantization parameter, the code amount is different when the quantization parameter is different or the code amount-encoding distortion function is different from the assumption. The correct comparison and evaluation of the conversion cost cannot be performed, and an appropriate prediction mode is not selected.

  An object of the present invention is to provide an image encoding method and apparatus capable of solving the above-described problems of the prior art, applying an intra-frame predictive encoding method, and performing encoding excellent in encoding efficiency and image quality. It is to provide.

  In order to achieve the above-described object, the present invention divides each frame of a moving image into a plurality of pixel blocks, and sequentially encodes each pixel block using an intra-frame prediction method. The following procedure is included.

  (1) A procedure for reading pixel blocks from a moving image in a predetermined order, a procedure for encoding the read pixel blocks in a plurality of prediction modes with different reference pixels, and a coding cost of each pixel block in the prediction mode The encoding procedure reflects the prediction result for each combination of prediction modes that can be selected in each preceding pixel block that affects the encoding cost of the pixel block to be encoded. In addition, prediction encoding is performed in a plurality of prediction modes, and further includes a procedure for determining a prediction mode for each of the preceding pixel blocks based on a combination of prediction modes that minimizes the sum of encoding costs.

  (2) The method further comprises a step of selecting a part of prediction modes with low coding cost based on the coding cost calculated for each prediction mode in each pixel block, and each pixel block has its own coding cost. For each combination of some prediction modes selected in each of the preceding pixel blocks that affect the prediction block, the prediction result is reflected and prediction encoding is performed in a plurality of prediction modes.

  (3) For each pixel block, among the combinations of prediction modes that can be selected in each preceding pixel block, prediction of the combination with the smallest total coding cost among the combinations with the same prediction mode in the immediately preceding pixel block Reflecting the result, the prediction encoding is performed in a plurality of prediction modes.

According to the present invention, the following effects are achieved.
(1) The prediction mode for intra-frame predictive encoding of each pixel block is not limited to the encoding cost of the pixel block, but the encoding cost of other pixel blocks that are affected by the encoding result of the pixel block. Therefore, the encoding cost for the entire frame can be kept low.
(2) Based on the coding cost calculated for each prediction mode in each pixel block, if some prediction modes with low coding cost are selected, prediction optimal for predictive coding of each pixel block The mode can be searched with a small amount of calculation.
(3) Among the combinations of prediction modes that can be selected by each pixel block in front of each pixel block, prediction of the combination having the smallest total coding cost among the combinations having the same prediction mode in the immediately preceding pixel block If the results are reflected and prediction coding is performed in a plurality of prediction modes, a prediction mode optimal for prediction coding of each pixel block can be searched with a small amount of calculation.

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the best embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a main part of a video encoding apparatus according to the present invention.

  The moving image signal S1 is temporarily stored in the buffer memory 1 and then used for intra-frame prediction by the intra-frame prediction unit 2 or used for inter-frame prediction by the inter-frame prediction unit 9. The prediction scheme switching unit 11 switches the prediction scheme on a frame basis based on the coding cost, connects the orthogonal transformation unit 3 to the intraframe prediction unit 2 for intraframe prediction, and subtracts the orthogonal transformation unit 3 for interframe prediction. 13 is connected.

  The motion prediction unit 14 generates a motion prediction vector based on the moving image signal S1. The inter-frame prediction unit 9 performs motion compensation prediction based on a motion vector between the motion prediction vector and the reference image signal stored in the frame memory 10, and creates a motion compensation prediction signal S2.

  In the subtracter 13, the motion compensated prediction signal S2 is subtracted from the moving image signal S1 to generate a prediction residual signal S3. The prediction residual signal S3 is orthogonally transformed by the orthogonal transform unit 3 in units of blocks having a certain size to become orthogonal transform coefficient information S4. The orthogonal transform coefficient information S4 is quantized by the quantization unit 4. The orthogonal transform coefficient information S5 quantized by the quantizing unit 4 is variable-length coded by the entropy coding unit 5 and stored in the buffer memory 6, while being dequantized by the inverse quantization unit 7. The output of the inverse quantization unit 7 is subjected to inverse orthogonal transform by the inverse orthogonal transform unit 8. That is, the inverse quantization unit 7 and the inverse orthogonal transform unit 8 perform processing reverse to that of the quantization unit 4 and the orthogonal transform unit 3, and the output of the inverse orthogonal transform unit 8 approximates the prediction residual signal S3. Signal is obtained.

  The output of the inverse orthogonal transform unit 8 is added to the prediction signal S2 output from the inter-frame prediction unit 9 by the adder circuit 12 to generate a local decoded signal S6. The local decoded signal S6 is stored in the frame memory 10 as a reference image signal after block noise is removed by the deblocking filter 15.

  FIG. 2 is a flowchart showing the operation of the first embodiment of the present invention, and mainly shows the operation of the intra-frame prediction unit 2.

  In step S1, one of the pixel blocks is acquired from the buffer memory 1 as the current target block TBn. This pixel block is a 16 × 16 pixel block or a 4 × 4 pixel block if the encoding target is a luminance signal. In step S2, any one of a plurality of applicable prediction modes is selected. That is, if the encoding target is a 16 × 16 pixel block, one of the four prediction modes described with reference to FIG. 8 is selected, and if the encoding target is a 4 × 4 pixel block, FIG. Any of the nine prediction modes described with respect to is selected. However, in the present embodiment, the selectable prediction modes are limited to three (M1, M2, M3) in advance, and description will be made assuming that any one of these is selected. Depending on the position of the target block, selectable prediction modes may be further limited.

  In step S3, the current prediction block TBn is subjected to intraframe prediction encoding by applying the current prediction mode selected in step S2. At this time, as will be described in detail later in this embodiment, the prediction modes of the previous attention block TBn-1 and the previous attention block TBn-2 that are positioned ahead of the current attention block TBn have not yet been determined. Only two prediction mode candidates that are advantageous from the viewpoint of encoding cost are selected. Therefore, in this attention block TBn, it is assumed that each of the previous and previous attention blocks TBn-1 and TBn-2 is encoded on one of the two prediction mode candidates, and on the other. The prediction calculation is performed in three prediction modes by reflecting each prediction result for each combination with the case where it is assumed.

  In step S <b> 4, the encoding cost calculation unit 201 calculates the encoding cost for each of the implemented prediction modes. As the coding cost, the amount of generated code and the estimation of the prediction error can be used alone or in combination. In step S5, it is determined whether or not predictive coding and coding cost calculation have been completed in all prepared prediction modes (in this embodiment, three types M1, M2, and M3). If not completed, the process returns to step S2, and the above-described processes are repeated while switching the prediction mode.

  As described above, when prediction encoding and encoding cost calculation are completed for all three prepared prediction modes (M1, M2, M3), the process proceeds to step S6, and a prediction mode with a low encoding cost is selected this time. As a prediction mode candidate of the block of interest, the prediction mode supplement selection unit 202 selects it. In the present embodiment, description will be made assuming that the top two prediction modes with the lowest coding costs are selected as candidates.

  In step S 7, the two selected prediction mode candidates (for example, M 1 and M 2) are stored in the prediction mode candidate storage unit 203. In step S8, it is determined whether or not selection of prediction mode candidates has been completed for a predetermined number (for example, three in this embodiment) of consecutive pixel blocks. If not completed, the process returns to step S1 to read the next pixel block, and the above-described processes are repeated for this pixel block to select prediction mode candidates.

  When selection of prediction mode candidates and cost calculation are completed for the three target blocks TBn-2, TBn-1, and TB (previous times, previous time, and current time), the process proceeds to step S9. In step S9, the prediction mode of the previous block of interest TBn-2 is determined by the prediction mode determination unit 204 based on the total encoding cost ΣC from the previous block of interest TBn-2 to the current block of interest TBn. The

  FIG. 3 is a diagram schematically representing a prediction mode candidate selection method by the prediction mode candidate selection unit 202 and a prediction mode determination method by the prediction mode determination unit 204.

  If two prediction mode candidates M1 and M2 are selected in the previous block of interest TBn-2 and their respective encoding costs are “1” and “2”, the next block adjacent to this block of interest TBn-2 is the next. In the previous target block TBn-1 to be predictively encoded, the results of the respective predictive encoding for the case in which the previous target block TBn-2 is encoded in the prediction mode M1 and the case in which the target block TBn-2 is encoded in the prediction mode M2 As a result, prediction encoding is performed in the three prediction modes (M1, M2, M3), and the top two prediction modes with the lowest encoding costs are selected as prediction mode candidates for the block of interest TBn-1.

  In the illustrated example, if the previous block of interest TBn-2 is encoded in the prediction mode M1, the prediction modes M1 and M3 are selected in the previous block of interest TBn-1, and the respective encoding costs are “6”, “ 5 ". If the previous block of interest TBn-2 is encoded in the prediction mode M2, the prediction modes M2 and M3 are selected in the previous block of interest TBn-1 and the encoding costs are “4” and “3”, respectively. .

  Further, in the current block of interest TBn that is next predictively encoded adjacent to the previous block of interest TBn-1, the previous blocks of interest TBn-2 and TBn-1 of the previous time and in the prediction mode M1 and M1, respectively. The prediction codes for all cases encoded, cases encoded in prediction modes M1 and M3, cases encoded in prediction modes M2 and M2, and cases encoded in prediction modes M2 and M3 Reflecting the result of encoding, predictive encoding is performed in the three prediction modes, and the top two prediction modes with the lowest encoding costs are selected.

  In the example shown in the figure, the prediction mode M2 and M1 are selected in the current block of interest TBn if the previous block of attention TBn-2 and TBn-1 are encoded with the combination of the prediction mode M1 → M1, respectively. On the other hand, if the encoding cost is “7” and “8”, respectively, if the prediction mode M1 → M3 is encoded, the prediction mode M3 and M1 are selected in the current block of interest TBn, and the encoding costs are respectively “9” and “10”. Similarly, the top two prediction mode candidates with the lowest coding costs are selected for each combination of prediction mode candidates in the previous block of interest TBn-2 and TBn-1 in the previous time and the previous time, respectively.

  As described above, when selection of prediction mode candidates and cost calculation are completed for each of the previous and current attention blocks TBn-2, TBn-1, and TBn, the previous attention block TBn-2 is obtained in step S9. The total encoding cost ΣC from the current block to the current block TBn is obtained for each combination of prediction mode candidates, and the prediction mode of the previous block of interest TBn-2 is finally determined based on the combination whose cost total ΣC shows the minimum value. The

  In the present embodiment, the total cost ΣC when the respective target blocks TBn-2, TBn-1, TBn of the previous time, the previous time, and the current time are encoded by the combination of the prediction modes M1 → M1 → M2 is “14”, The total cost ΣC when encoded in a combination of prediction modes M1 → M1 → M1 is “15”, and the total cost ΣC when encoded in a combination of prediction modes M2 → M3 → M2 is “16”. . Since the minimum value of the total cost ΣC is “14” in the combination of the prediction modes M1 → M1 → M2, the prediction mode of the previous block of interest TBn-2 is determined as its own prediction mode M1 in this combination.

  Returning to FIG. 2, in step S <b> 10, it is determined whether or not the prediction mode of all the pixel blocks has been determined, and if not completed, the process returns to step S <b> 1. In step S1, the next pixel block TBn-1 of the pixel block TBn-2 for which the prediction mode has been determined is newly acquired as the current block of interest. That is, if the prediction mode of the previous block of interest TBn-2 is determined as described above, the previous block of interest TBn-1 is acquired, and the above-described processes are repeated in order from this pixel block (block of interest). . Then, the prediction mode of the previous block of interest TBn-1 is determined based on the encoding cost in the previous, current and next blocks of interest TBn-1, TBn, TBn + 1.

  FIG. 4 is a diagram schematically illustrating a method for determining the prediction mode of each pixel block in the present embodiment. Based on the encoding cost from the pixel block TBn-2 to TBn, the first pixel block TBn-2 is determined. The prediction mode is determined, the prediction mode of the first pixel block TBn-1 is determined based on the encoding cost from TBn-1 to TBn + 1, and the leading mode is determined based on the encoding cost from TBn to TBn + 2. The prediction mode of the pixel block TB is determined

  In the embodiment described above, it is assumed that only the prediction mode of the first pixel block is determined based on the sum ΣC of encoding costs of N pixel blocks (three in this embodiment). However, the present invention is not limited to this, and the prediction mode of the first M (<N) pixel blocks is determined based on the sum of the encoding costs of N pixel blocks, This may be repeated. FIG. 5 shows an example in which the prediction modes of the two first pixel blocks are determined based on the sum of coding costs in three consecutive pixel blocks.

  Alternatively, the prediction modes of all N pixel blocks may be determined based on the total encoding cost of N pixel blocks, and this may be repeated. FIG. 6 shows an example in which prediction modes for all three pixel blocks are determined based on the sum of coding costs in three consecutive pixel blocks.

  Furthermore, in the above-described embodiment, only the upper prediction mode with the lowest coding cost is selected from the selectable prediction modes, and the prediction mode of each pixel block behind is selected with respect to the selected prediction mode. Although described as an example, all selectable prediction modes may be evaluated.

  According to the present embodiment, the prediction mode for intra-frame prediction of each pixel block is not limited to the encoding cost of the pixel block alone, but other pixel blocks that are affected by the prediction encoding result of the pixel block. Therefore, the coding cost for the entire frame can be kept low.

  By the way, as in the first embodiment described above, each pixel block performs predictive coding for all combinations of prediction modes that can be selected by each pixel block ahead of itself, and calculates the sum of the coding costs. As the number of pixel blocks for obtaining the sum increases, the processing load of the intra-frame prediction unit 2 increases and the calculation time becomes longer. Therefore, in the second embodiment of the present invention described below, the calculation amount of the encoding cost is limited in each pixel block.

  FIG. 7 is a diagram schematically showing the operation of the second embodiment for reducing the amount of calculation, and here again, the case where there are three types of prediction modes (M1, M2, M3) that can be selected is taken as an example. explain.

  In the pixel block TBn-1, the prediction results for the case where the immediately preceding pixel block TBn-2 is encoded in the prediction mode M1, the case of prediction in the prediction mode M2, and the case of encoding in the prediction mode M3 are shown. Reflecting, predictive encoding is performed in the three prediction modes M1, M2, and M3, and respective encoding costs are obtained. That is, the prediction calculation is executed only 9 times.

  Further, in the next pixel block TBn adjacent to this pixel block TBn-1, the combinations having the same prediction mode in the immediately preceding pixel block TBn-1 compete with each other, and only for the combination with the minimum coding cost. Reflecting the prediction result, the prediction calculation is performed in the three prediction modes M1, M2, and M3, and the other combinations are discarded.

  That is, the prediction mode of the pixel block TBn-1 is M1 because the prediction modes of the pixel blocks TBn-2 and TBn-1 are three combinations of M1 → M1, M2 → M1, and M3 → M1. So, if these three sets of coding costs are compared, and the combination of M2 → M1, for example, is the minimum cost, the prediction calculation will be performed in the three prediction modes M1, M2, and M3, reflecting the prediction result of this combination. The encoding cost is obtained only for the combination of M2, M1, M1, M2, M1, M2, and M2, M1, M3.

  Similarly, the prediction mode of the pixel block TBn-1 is M2 because the prediction modes of the pixel blocks TBn-2 and TBn-1 are a combination of M1 → M2, a combination of M2 → M2, and a combination of M3 → M2. Therefore, if the combination of M2 → M2 is the minimum cost, for example, if the combination of M2 → M2 is the minimum cost, predictive encoding is performed in the three prediction modes M1, M2, and M3 reflecting the prediction result of this combination. The encoding cost is obtained only for the combination of M2 → M2 → M1, the combination of M2 → M2 → M2, and the combination of M2 → M2 → M3.

  Similarly, the prediction mode of the pixel block TBn-1 is M3 because the prediction modes of the pixel blocks TBn-2 and TBn-1 are a combination of M1 → M3, a combination of M2 → M3, and a combination of M3 → M3. Therefore, if the combination of M3 → M3 is the minimum cost, for example, if the combination of M3 → M3 is the minimum cost, the prediction encoding by this combination is reflected and the prediction encoding is performed in the three prediction modes M1, M2, and M3. The encoding cost is obtained only for the combination of M3 → M3 → M1, the combination of M3 → M3 → M2, and the combination of M3 → M3 → M3.

  Similarly, the prediction mode of the pixel block TBn + 1 is M1, M2, M1, M2, M2, M1, and M3, M3, M1, and so on for the next pixel block TBn + 1. For example, if the combination of M2 → M2 → M1 is a minimum cost, predictive encoding is performed in three prediction modes M1, M2, and M3 reflecting the prediction result of this combination, and M2 → Coding costs are obtained only for the combinations of M2 → M1 → M1, M2 → M2 → M1 → M2, and M2 → M2 → M1 → M3.

  Similarly, the prediction mode of the pixel block TBn is M2, since M2 → M1 → M2, M2 → M2 → M2, and M3 → M3 → M2 are evaluated. If the combination of M2 → M1 → M2 is the minimum cost, the prediction results of this combination are reflected and prediction coding is performed in three prediction modes M1, M2, M3, and the combination of M2 → M1 → M2 → M1 The encoding cost is obtained only for the combination of M2 → M1 → M2 → M2 and the combination of M2 → M1 → M2 → M3.

  Similarly, the prediction mode of the pixel block TBn is M3 because there are three sets of M2 → M1 → M3, M2 → M2 → M3, and M3 → M3 → M3. If the combination of M2 → M1 → M3 is the minimum cost, the prediction results of this combination are reflected and prediction encoding is performed in the three prediction modes M1, M2, M3, and the combination of M2 → M1 → M3 → M1. The encoding cost is obtained only for the combination of M2 → M1 → M3 → M2 and the combination of M2 → M1 → M3 → M3.

  As described above, when the sum of the coding costs of the pixel blocks of a predetermined reference fraction is obtained, the prediction mode of each pixel block is determined by the same method as in the first embodiment based on the combination of the minimum costs. Is done.

  According to the present embodiment, the total number of coding cost calculations for each block of interest is limited to nine times regardless of the number of pixel blocks to which the coding cost is referenced. Can be encoded and compressed.

  In the above-described embodiment, the present invention has been described with an example of prediction in units of 16 × 16 pixel macroblocks. However, the present invention is not limited to this, and each of the embodiments described with reference to FIG. The same applies to 4 × 4 prediction within a macroblock.

  Also, in H.264 / AVC, the coding cost by intra-frame prediction and the coding cost by inter-frame prediction are compared for each macroblock, and a prediction method with low coding cost is adopted. The embodiment described above can be applied not only at the time of encoding but also at the time of calculating the encoding cost of intra-frame prediction when determining the prediction method.

It is the block diagram which showed the structure of the principal part of the moving image encoder which concerns on this invention. It is the flowchart which showed operation | movement of 1st Embodiment of this invention. It is the figure which expressed typically the selection method and determination method of a prediction mode candidate. It is the figure which expressed typically the determination method of the prediction mode of each pixel block. It is the figure which showed the method of determining the prediction mode of the earliest two pixel blocks based on the sum total of the encoding cost in three continuous pixel blocks. It is the figure which showed the method of determining the prediction mode of all the three pixel blocks based on the sum total of the encoding cost in three continuous pixel blocks. It is the figure which expressed typically the operation | movement of 2nd Embodiment which reduces the amount of calculations. It is a figure for demonstrating the prediction mode of 16 * 16 prediction regarding the luminance signal of H.264 / AVC. It is a figure for demonstrating the prediction mode of 4x4 prediction regarding the luminance signal of H.264 / AVC. It is a figure for demonstrating the prediction mode of 8 pixels x 8 lines in the macroblock regarding the color difference signal of H.264 / AVC.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Buffer memory, 2 ... Intra-frame prediction part, 3 ... Orthogonal transformation part, 4 ... Quantization part, 5 ... Entropy encoding part, 6 ... Buffer memory, 7 ... Inverse quantization part, 8 ... Inverse orthogonal transformation part, DESCRIPTION OF SYMBOLS 9 ... Inter-frame prediction part, 10 ... Frame memory, 11 ... Prediction system switching part, 12 ... Adder circuit, 13 ... Subtractor, 14 ... Motion prediction part, 15 ... Deblocking filter

Claims (9)

In a predictive encoding method of a moving image, each frame of a moving image is divided into a plurality of pixel blocks, and each pixel block is sequentially encoded by an intra-frame prediction method.
A procedure for reading out pixel blocks from a moving image in a predetermined order;
A procedure for encoding the read pixel block in a plurality of prediction modes with different reference pixels;
And a procedure for calculating the encoding cost of each pixel block for each prediction mode,
In the encoding procedure, for each combination of prediction modes that can be selected in each preceding pixel block that affects the encoding cost of the pixel block to be encoded, a prediction code is reflected in a plurality of prediction modes by reflecting the prediction result. Further,
And a procedure for determining a prediction mode for each of the preceding pixel blocks based on a combination of prediction modes that minimizes the total encoding cost.   In the procedure of determining the prediction mode, the prediction mode of M (<N) pixel blocks is determined from the front based on the combination of prediction modes of N pixel blocks that minimize the sum of coding costs. The moving picture predictive coding method according to claim 1, wherein:   In the procedure of determining the prediction mode, the prediction mode of all N pixel blocks is determined based on a combination of prediction modes of N pixel blocks that minimizes the sum of encoding costs. The moving image predictive encoding method according to claim 1. Further comprising a procedure for selecting a part of the prediction modes with a low coding cost based on the coding cost calculated for each prediction mode in each pixel block;
Each pixel block is predictively encoded in a plurality of prediction modes by reflecting the prediction result for each combination of some prediction modes selected in each preceding pixel block that affects its own encoding cost. The moving picture predictive coding method according to claim 1, wherein:   Each pixel block reflects the prediction result of the combination of the prediction modes that can be selected in each pixel block in front of the combination that has the same prediction mode in the immediately preceding pixel block and that has the lowest total coding cost. 4. The moving picture predictive coding method according to claim 1, wherein the predictive coding is performed in a plurality of prediction modes. In a predictive encoding apparatus for a moving image, in which each frame of a moving image is divided into a plurality of pixel blocks, and each pixel block is sequentially encoded by an intra-frame prediction method.
Means for reading out pixel blocks from a moving image in a predetermined order;
Means for encoding the read pixel block in a plurality of prediction modes with different reference pixels;
Means for calculating the encoding cost of each pixel block for each prediction mode,
In the encoding means, for each combination of prediction modes that can be selected in each preceding pixel block that affects the encoding cost of the pixel block to be encoded, a prediction code is reflected in a plurality of prediction modes by reflecting the prediction result. Further,
And a means for determining a prediction mode for each of the preceding pixel blocks based on a combination of prediction modes that minimizes the sum of encoding costs.   The means for determining the prediction mode determines the prediction mode of M (<N) pixel blocks from the front based on a combination of prediction modes of N pixel blocks that minimize the sum of coding costs. The moving picture predictive coding apparatus according to claim 6.   The means for determining the prediction mode is characterized in that the prediction modes of all N pixel blocks are determined based on a combination of prediction modes of N pixel blocks that minimize the sum of encoding costs. The moving image predictive coding apparatus according to claim 6. Based on the encoding cost calculated for each prediction mode in each pixel block, further comprising means for selecting some prediction modes with low encoding costs,
Each pixel block is predictively encoded in a plurality of prediction modes by reflecting the prediction result for each combination of some prediction modes selected in each preceding pixel block that affects its own encoding cost. The moving picture predictive coding apparatus according to any one of claims 6 to 8.

Images