JPWO2012141221A1 - Moving picture coding apparatus, moving picture coding method, moving picture coding program, and computer-readable recording medium - Google Patents

Abstract

An object of the present invention is to increase the efficiency of inter prediction encoding and reduce the data size. One or more candidate criteria having high relevance in the plurality of blocks from among a plurality of blocks existing at a position corresponding to the sub-block and its surroundings in a previously encoded previous image frame A step of extracting a block, a step of selecting a block closest to the encoding target sub-block as a reference block from among the extracted one or more candidate reference blocks, and a plurality of pixels constituting the selected reference block Generating a second reference pixel according to a plurality of pre-defined additional prediction modes based on the value, generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel, Performing rate distortion optimization based on the two predicted pixels. [Selection] Figure 19

Description

  The present invention relates to a moving image encoding apparatus, a moving image encoding method, a moving image encoding program, and a computer-readable recording medium used for encoding and decoding moving images.

  In recent years, high-definition moving images such as DVD and terrestrial digital high-definition have been rapidly spreading, and the amount of information has increased dramatically due to the high bit rate and high resolution of moving images. It is essential. As an encoding method for performing such data compression, currently H.264 is used. H.264 / AVC is prevalent. H. H.264 / AVC is a video coding standard recommended by ITU-T and others in 2003, and uses a conventional hybrid coding structure based on Motion Compensation (MC) DCT. H. In order to reduce temporal and spatial redundancy in H.264 / AVC, intra prediction encoding (INTRA prediction encoding; hereinafter also referred to as “intra prediction”) and inter prediction encoding (INTER) are provided as two types of prediction modes. Prediction coding, interframe prediction; hereinafter also referred to as “inter prediction”).

  Inter prediction is a prediction method that takes a difference between corresponding images between two frames having different times, and aims to reduce temporal redundancy. Here, in order to effectively perform motion compensation, a 16 × 16 pixel (pixel) block (Macroblock: hereinafter referred to as “macroblock”) can be encoded with 7 types of block sizes of 4 × 4 to 16 × 16 pixels. In addition, in order to further improve the coding efficiency, a plurality of frames and 1/2, 1/4 accuracy prediction can also be used.

  However, these motion compensation tools can only reduce temporal redundancy for continuous motion movies. That is, sufficient encoding efficiency cannot be achieved by inter prediction for a moving image with a large change in motion. In such a case, the intra prediction mode is used to reduce the spatial redundancy. H. In H.264 / AVC, 4 × 4 or 16 × 16 pixel mode is used for prediction from adjacent blocks. These two modes are pre-encoding by rate-distortion optimization (RDO) processing, and the most efficient mode that makes the final encoding mode appropriate is selected. In most cases, this method works well because intra mode works well as a supplemental tool when inter mode does not provide good coding.

  However, depending on the moving image, there is a case where the redundancy cannot be reduced from either viewpoint of temporal redundancy and spatial redundancy. Even in such a case, H.C. In H.264 / AVC, either inter prediction or intra prediction, the bit rate and image quality are relatively the best, and the mode with low RDC (Rate-Distortion Cost: hereinafter referred to as “rate distortion cost” or simply “cost”) is used. As a result, the bit rate increases. An example of such a moving image is “football”. In a moving image obtained by shooting a football game, it is difficult to reduce temporal redundancy due to the quick movement of the players, and it is difficult to estimate the spatial redundancy due to extremely complicated content. As a result, there is a problem that the bit rate becomes unstable, and the mounting specifications required for hardware and software are increased accordingly.

  On the other hand, a multi-core CPU and a high-performance GPU architecture are being newly developed in order to meet such demands for increasing hardware computing power. For example, multi-core arithmetic changes the basic concept of many fields including multimedia processing. However, the use of such techniques merely concentrates all conventional video encoding tools on the encoding process. As a result, the fundamental problem that parallel processing becomes difficult due to complex correlation between adjacent blocks cannot be solved.

  On the other hand, as a next-generation image compression method, H.264 is used. H.265 / AVC (HVC: also called High Efficiency Video Coding, etc.) is being developed. The goal is to achieve approximately twice the coding efficiency over H.264 / AVC. However, it is not easy to achieve such a high compression ratio, and many of the currently proposed methods take an approach to increase the total compression ratio by combining multiple existing or new compression methods. It has been. In other words, it is not easy to achieve a high compression rate with a single encoding method (for example, Patent Document 1 and Non-Patent Documents 1 to 9).

Japanese Patent Laid-Open No. 2005-13003

"IT coding of audiovisual objects-Part 10: Advanced video coding" ISO / IEC 14496-10: 2003 (December 2003) (ISO / IEC 14496-10: 2003. Information technology coding of audio-visual objects-Part 10: advanced video coding. Dec. 2003.) "Computer Integrated Device Architecture Programming Guide, Version 2.3, 2009" Nvidia (NVIDIA, NVIDIA CUDA Compute Unified Device Architecture Programming Guide Version 2.3, 2009.) T.A. D. Chang, Y. H. Chen, C.H. H. Tsai, Y. J. et al. Chen, L. G. Chen “H.264 / AVC High Profile Intra Prediction Algorithm and Architecture Design” Image Coding Symposium Proceedings (2007) (TD Chuang, YH Chen, CH Tsai, YJ Chen, and LG Chen, “Algorithm and architecture design for (intra prediction in H.264 / AVC high profile ", in Proc. of Picture Coding Symposium, 2007.) L. Chang, S. W. Ma, W. Gao "Position-Dependent Linear Intra Prediction for Image Coding" IEEE 17th International Conference on Image Processing (ICIP) Proceedings 2877-2880 (2010) (L. Zhang, SWMa, W. Gao, "Position dependent linear intra prediction fro image coding ", in Proc. of IEEE 17th International Conference on Image Processing (ICIP), pp.2877-2880, 2010.) JM14.2 <http://iphome.hhi.de/suehring/tml/> Satoshi Okubo et al. “Revised Third Edition H.264 / AVC Textbook” Impress "Patent Map by Technology Field Electricity 14 Digital Video Compression Technology" Japan Patent Office JVT of ISO / IEC MPEG & ITU-T VCEG, "H.264 / MPEG-4 AVC Reference Software Manual" July, 2007 T. Wiegand et al, "Overview of the H.264 / AVC Video Coding Standard" IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 13, NO. 7, JULY 2003

  The present invention has been made in view of such a background. The main object of the present invention is to provide a more efficient moving picture coding apparatus, moving picture coding method, moving picture coding program, and computer-readable recording medium.

Means for Solving the Problems and Effects of the Invention

  In order to achieve the above object, a first moving image encoding method of the present invention is a moving image encoding method for encoding moving image data, the step of acquiring moving image data, and the acquired When encoding an arbitrary macroblock constituting an arbitrary current image frame of moving image data for each subblock constituting the macroblock, it corresponds to the subblock in the previous image frame that has already been encoded. A step of extracting one or more candidate reference blocks having high relevance in the plurality of blocks from the plurality of blocks existing in the position and the periphery thereof, and a code from the extracted one or more candidate reference blocks A step of selecting a block closest to the sub-block to be converted as a reference block, and a plurality of pixel values constituting the selected reference block. Generating a second reference pixel according to a plurality of additional prediction modes, generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel, and based on the second prediction pixel , Performing rate distortion optimization, selecting an intra prediction mode with the lowest rate distortion cost, and encoding according to the selected intra prediction mode. As a result, it is possible to realize more efficient encoding without substantially increasing the amount of calculation, that is, without changing the conventional hardware.

  In the second moving image encoding method, in the step of generating the second prediction pixel, the prediction pixel acquired according to a predetermined intra prediction mode is generated, and at the time of the rate distortion optimization, the second prediction pixel is generated. In addition, the intra prediction mode with the lowest rate distortion cost can be selected including the prediction pixels. Thereby, since the temporal correlation of the previous image frame can be used in encoding in the intra prediction mode, encoding with a data amount smaller than that in the conventional intra prediction mode can be realized. In addition, since the same algorithm as that of the existing intra prediction mode is adopted, existing hardware can be used, and there is almost no cost penalty for introduction, and it can be implemented at low cost on existing equipment.

  Furthermore, in the third moving image encoding method, the step of extracting the one or more candidate reference blocks is an intra prediction mode that is most frequently used among a plurality of intra prediction modes employed in encoding the plurality of blocks. The prediction mode can be selected as the most numerous mode, and all the blocks adopting the most frequent mode can be extracted as candidate reference blocks.

  Furthermore, in the fourth moving image encoding method, the step of generating the second reference pixel adds pixels arranged in the vertical direction among a plurality of pixel values constituting the selected reference block, The average value divided by the number of pixels can be used as the pixel value of the second reference pixel.

  Furthermore, in the fifth moving image encoding method, the step of generating the second reference pixel adds pixels aligned in the horizontal direction among a plurality of pixel values constituting the selected reference block, The average value divided by the number of pixels can be used as the pixel value of the second reference pixel.

  Furthermore, in the sixth moving image encoding method, the step of generating the second reference pixel is an average obtained by adding all of a plurality of pixel values constituting the selected reference block and dividing by the number of all pixels. The value can be the pixel value of the second reference pixel.

  Furthermore, according to a seventh video encoding method, the plurality of additional prediction modes are H.264 and H.264. Inverse calculation of four intra prediction modes of 4 × 4 blocks of luminance components defined by the H.264 / AVC standard.

  Furthermore, in the eighth moving image encoding method, the plurality of additional prediction modes are H.264 and H.264. Inverse calculation of four intra prediction modes of 16 × 16 blocks of luminance components defined by the H.264 / AVC standard.

  Furthermore, the ninth video coding method adopts the inter prediction mode preferentially over the intra prediction mode in encoding, and the intra prediction mode is selected when the inter prediction mode has a higher rate distortion cost than the intra prediction mode. Can be adopted.

  Furthermore, the tenth moving image encoding method can use original pixel data as image data to be encoded. Thereby, it is possible to perform parallel encoding based on macroblocks while suppressing an increase in the amount of calculation.

  Furthermore, in the eleventh moving image encoding method, the step of generating the second reference pixel weights the plurality of pixel values constituting the selected reference block in accordance with the distance from the target pixel. The step of generating the third reference pixel can be performed. Thereby, since the prediction which attached importance to the pixel with a short distance can be performed, a more exact reference pixel can be obtained.

  Furthermore, in the twelfth moving image encoding method, in the step of selecting the reference block, the head block is set as a key block for each of the sub-blocks, and position information is set for the key block. The blocks other than the key block can be substituted with the position information of the key block. Thereby, while providing position information to the inner key block of each reference block, it is possible to suppress an increase in the number of bits by using the position information of the key block for adjacent blocks.

  Furthermore, the thirteenth moving image encoding apparatus comprises moving image input means for acquiring moving image data, compression means for compressing moving image data input by the moving image input means, and the compression means. And a quantization means for quantizing the compressed data compressed in step (b), wherein the compression means is an arbitrary macro that constitutes an arbitrary current image frame of the acquired moving image data. In encoding a block for each sub-block constituting the macro block, a position corresponding to the sub-block in a previously encoded image frame and a plurality of blocks existing in the vicinity thereof are Extraction means for extracting one or more candidate reference blocks having high relevance among a plurality of blocks, and one or more candidate reference blocks extracted by the extraction means, In accordance with a plurality of pre-defined additional prediction modes based on a plurality of pixel values constituting a reference block selected by the reference block selection unit Second reference pixel generation means for generating a second reference pixel; and second reference prediction pixel generation means for generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel, Based on the second prediction pixel generated by the second reference prediction pixel generation means, rate distortion optimization can be executed and an intra prediction mode with the lowest rate distortion cost can be selected. As a result, it is possible to realize more efficient encoding without substantially increasing the amount of calculation, that is, without changing the conventional hardware.

  Further, the fourteenth moving image encoding program is a moving image encoding program for encoding moving image data. The fourteenth moving image encoding program has a function of acquiring moving image data in a computer and an arbitrary present of the acquired moving image data. When an arbitrary macroblock constituting an image frame is encoded for each subblock constituting the macroblock, the macroblock is present at a position corresponding to the subblock and the surrounding area in a previously encoded image frame. A function of extracting one or more candidate reference blocks having high relevance in the plurality of blocks, and the sub-block to be encoded is most frequently selected from the extracted one or more candidate reference blocks. Based on the function of selecting an approximate block as a reference block and a plurality of pixel values constituting the selected reference block, A function of generating a second reference pixel according to the plurality of additional prediction modes performed, a function of generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel, and a function of the second prediction pixel Thus, it is possible to realize a function of performing rate distortion optimization and selecting an intra prediction mode with the lowest rate distortion cost and a function of performing encoding according to the selected intra prediction mode. As a result, it is possible to realize more efficient encoding without substantially increasing the amount of calculation, that is, without changing the conventional hardware.

  Furthermore, a fifteenth computer-readable recording medium or recorded device stores the program. Recording media can store CD-ROM, DVD, Blu-ray (registered trademark), flexible disk, optical disk such as MO, magnetic disk, magneto-optical disk, magnetic medium such as magnetic tape, semiconductor memory, and other programs. Media. The program includes a program distributed in a download manner through a network line such as the Internet, in addition to a program stored and distributed in the recording medium. Further, the recorded devices include general-purpose or dedicated devices in which the program is implemented in a state where it can be executed in the form of software, firmware, or the like. Furthermore, each process and function included in the program may be executed by program software that can be executed by a computer, and each part of the process and function may be performed with hardware such as a predetermined gate array (FPGA, ASIC), or program software. You may implement | achieve in the format with which the partial hardware module which implement | achieves some elements of hardware is mixed.

It is a block diagram which shows the moving image encoder which concerns on one embodiment of this invention. It is a schematic diagram which shows the example of the macroblock in inter prediction. It is a schematic diagram which shows the macroblock of 16 * 16 intra prediction mode. FIG. 4 is a schematic diagram showing prediction modes 0 to 3 in FIG. 3, in which FIG. 4A is mode 0, FIG. 4B is mode 1, FIG. 4C is mode 2, and FIG. Respectively. It is a schematic diagram which shows the macroblock of 4x4 intra prediction mode. 6 is a schematic diagram illustrating prediction modes 0 to 8 in FIG. 5, FIG. 6A is mode 0, FIG. 6B is mode 1, FIG. 6C is mode 2, and FIG. 6D is mode 3. 6 (e) is mode 4, FIG. 6 (f) is mode 5, FIG. 6 (g) is mode 6, FIG. 6 (h) is mode 7, FIG. 6 (i) is mode 8, and FIG. j) is a schematic diagram showing the pixel value of each pixel. FIGS. 7A and 7C show continuous frame images in the moving image, and FIGS. 7B and 7D show the encoding methods of the frame images in FIGS. 7A and 7C, respectively. It is an image figure. FIG. 8A is an image diagram showing a frame image encoding method in frame n-1, and FIG. 8B is an image diagram showing a frame image encoding method in frame n. It is a schematic diagram which shows a mode that intra prediction is selected before and behind a frame image. It is a schematic diagram which shows the pixel which comprises the macroblock of encoding object. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the moving image of a bus | bath. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the moving image of a coast guard. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the moving image of football. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the foreman's moving image. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the moving image of a mobile. It is a graph which shows the simulation result which compared the efficiency with the original pixel and encoding pixel in the moving image of a tamppet. It is a flowchart which shows the procedure of temporal prediction intra mode. It is a schematic diagram which shows a mode that the reference pixel used as a candidate is selected. It is a schematic diagram which shows the prediction mode added by the temporal prediction intra mode, FIG.19 (a) is addition mode 0, FIG.19 (b) is addition mode 1, FIG.19 (c) is addition mode 2, FIG. FIG. 19E shows the addition mode 4, FIG. 19F shows the addition mode 5, FIG. 19G shows the addition mode 6, FIG. 19H shows the addition mode 7, and FIG. (I) is a schematic diagram showing the add mode 8, and FIG. 19 (j) is a schematic diagram showing the pixel value of each pixel. It is a graph which shows the simulation result which compared the efficiency with the conventional encoding in the moving image of a bus | bath, and the encoding which concerns on an Example. It is a graph which shows the simulation result which compared the efficiency by the conventional encoding in the moving image of a coast guard, and the encoding which concerns on an Example. It is a graph which shows the simulation result which compared the efficiency by the conventional encoding in the moving image of football, and the encoding which concerns on an Example. It is a graph which shows the simulation result which compared efficiency with the conventional encoding in the foreman moving image, and the encoding which concerns on an Example. It is a graph which shows the simulation result which compared efficiency with the conventional encoding in the moving image of a mobile, and the encoding which concerns on an Example. It is a graph which shows the simulation result which compared efficiency with the conventional encoding in the moving image of a tampet, and the encoding which concerns on an Example. It is an enlarged view which shows the candidate reference | standard block of additional mode x in the reference | standard macroblock of previous frame n-1. FIG. 27 is a schematic diagram showing a state in which a block most similar to the current macro block is searched from the candidate reference blocks of FIG. 26. It is a schematic diagram which shows the position which a reference | standard block can take in a macroblock. It is a schematic diagram which shows a mode that the number of generated bits increases by adding position information. It is a schematic diagram which shows the example which matched the positional information in the macroblock of the frame n-1 and the frame n 1: 1. It is a schematic diagram which shows a mode that a key block is set with respect to a subblock. It is a graph which shows the result of having calculated the ratio with which the position of a reference block and another block is the same per subblock unit with respect to each sequence. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the moving image of a bus | bath, and the encoding of the comparative example 1. FIG. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the moving image of a coast guard, and the encoding of the comparative example 1. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the moving image of football, and the encoding of the comparative example 1. FIG. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the Forman's moving image, and the encoding of the comparative example 1. FIG. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the moving image of a mobile, and the encoding of the comparative example 1. FIG. It is a graph which shows the simulation result which compared the efficiency by the encoding which concerns on Example 2 in the moving image of a tampet, and the encoding of the comparative example 1. FIG.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the embodiments described below exemplify a moving image encoding device, a moving image encoding method, a moving image encoding program, and a computer-readable recording medium for embodying the technical idea of the present invention. Therefore, the present invention does not specify a moving image encoding apparatus, a moving image encoding method, a moving image encoding program, and a computer-readable recording medium as follows. Further, the present specification by no means specifies the members shown in the claims to the members of the embodiments. In particular, the dimensions, materials, shapes, relative arrangements, and the like of the component parts described in the embodiments are not intended to limit the scope of the present invention unless otherwise specified, and are merely explanations. It is just an example. Note that the size, positional relationship, and the like of the members shown in each drawing may be exaggerated for clarity of explanation. Furthermore, in the following description, the same name and symbol indicate the same or the same members, and detailed description thereof will be omitted as appropriate. Furthermore, each element constituting the present invention may be configured such that a plurality of elements are constituted by the same member and the plurality of elements are shared by one member, and conversely, the function of one member is constituted by a plurality of members. It can also be realized by sharing.

In this specification, connection to a computer, printer, external storage device and other peripheral devices for operation, control, input / output, display, and other processing connected to the moving image encoding apparatus is, for example, IEEE 1394, RS- 232x, RS-422, RS-423, RS-485, serial connection such as USB, parallel connection, or electrically connected via a network such as 10BASE-T, 100BASE-TX, 1000BASE-T . The connection is not limited to a physical connection using a wire, but may be a wireless connection using a wireless LAN such as IEEE802.1x, OFDM, or LTE, a radio wave such as Bluetooth (registered trademark), infrared light, optical communication, or the like. Further, a memory card, a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like can be used as a recording medium for storing text and image data to be searched, database construction, setting for searching, and the like.
(Moving picture encoding device)

  In FIG. 1 is a block diagram illustrating an example of a moving image encoding apparatus 100 that obtains a bit stream from image data by H.264 / AVC encoding processing. The moving image encoding apparatus 100 includes a motion estimation unit 102, an inter prediction unit 104, an intra prediction unit 106, a conversion unit 108, a quantization unit 110, an entropy encoding unit 114, an inverse quantization unit 116, an inverse conversion unit 118, and an RDO. A processing unit 119, a filter 120, and a frame memory 122 are provided. This moving image encoding apparatus 100 encodes input moving image data by compressing the data size. The compression work for reducing the data size is also called prediction because it is work for predicting the data of the block from the surrounding data and the previous frame. The input moving image data is composed of a plurality of pictures arranged on the time axis, and the picture is composed of a plurality of blocks. The block includes a macroblock and a subblock obtained by dividing the macroblock into two or four in the vertical or horizontal direction. The inter prediction unit 104 performs motion compensation. Here, one of the inter prediction unit 104 and the intra prediction unit 106 is selected by the RDO processing unit 119.

The moving picture coding apparatus 100 selects one prediction mode from among a plurality of coding modes (prediction modes) including inter prediction and intra prediction, and in the selected prediction mode, the macro coding of the current picture is selected. Encoding. Here, the moving picture coding apparatus 100 performs coding in all the prediction modes included in the inter prediction and the intra prediction, calculates the cost, determines the prediction mode with the smallest cost as the optimum mode, and performs this prediction. Encode in mode. Note that the selection and determination of the prediction mode is generally based on the coding cost. A function for evaluating the coding cost can be selected as appropriate. For example, there is a case where a method of predicting the number of generated bits is used without performing complete provisional encoding in order to reduce the complexity of evaluating the encoding cost.
(Inter prediction)

Inter prediction is performed by the motion estimation unit 102 and the inter prediction unit 104. The motion estimation unit 102 searches the reference picture for the prediction value of the macroblock of the current picture. In addition, when the reference block is searched in units of 1/2 pixel or 1/4 pixel, the inter prediction unit 104 calculates the intermediate pixel value and determines the reference block data value.
(Intra prediction)

  On the other hand, the intra prediction unit 106 performs intra prediction for searching for a prediction value of a macroblock of the current picture in the current picture. Whether to perform inter prediction or intra prediction on the current macroblock is determined after calculating the costs of all prediction modes. That is, by comparing the costs, the mode having the smallest value is determined as the optimum prediction mode of this block, and the macroblock is encoded in this mode.

  When inter prediction or intra prediction is performed and prediction data referred to by the macroblock of the current frame is searched, the prediction data is subtracted from the macroblock of the current picture, converted by the conversion unit 108, and converted by the quantization unit 110. Perform quantization. In order to reduce the amount of data at the time of encoding, a difference value obtained by subtracting a motion estimated reference block from a macroblock of the current frame is encoded. The quantized difference value is encoded by the entropy encoding unit 114.

  On the other hand, in order to obtain a reference picture used for inter prediction, the current picture is restored by passing the quantized picture through the inverse quantization unit 116 and the inverse transform unit 118. The current picture restored in this way is stored in the frame memory 122 and used for inter prediction for subsequent pictures. By passing the restored picture through the filter 120, a picture including some coding errors in the original picture can be obtained.

FIG. 2 shows variable blocks adopted by the macroblock during inter prediction. H. In inter prediction in H.264 / AVC, one 16 × 16 macroblock is divided into 16 × 16, 16 × 8, 8 × 16, or 8 × 8 blocks. Each 8 × 8 block is further divided into small units of 8 × 4, 4 × 8, and 4 × 4 blocks. Motion estimation and compensation are performed on each of the sub-blocks thus divided, and a motion vector is determined. By performing inter prediction using such various types of variable blocks, it is possible to perform effective coding according to the characteristics and motion of moving images.
(Inter prediction unit 104)

The inter prediction in FIG. 1 is executed by the inter prediction unit 104. Inter-prediction is also called inter-frame prediction, and is a prediction method that takes a difference between corresponding images between two frames having different times, and aims to reduce temporal redundancy. In inter-prediction coding, conventionally, prediction of 8 × 8 pixel block size or larger has been used. In H.264 / AVC, prediction with a 4 × 4 pixel block size is possible, and more accurate prediction than in the conventional method is possible by motion compensation from a plurality of already encoded reference images. As described above, the number of prediction modes that can be selected for each block is increased, and a prediction mode with higher prediction efficiency is selected to improve the encoding efficiency.
(Intra prediction mode)

  Next, details of the conventional intra prediction mode will be described with reference to FIGS. H. In the H.264 / AVC standard, four different intra prediction modes for 16 × 16 blocks as shown in FIGS. 3 and 4, and nine different intra prediction modes for 4 × 4 blocks as shown in FIGS. Is available. 4 (a) to 4 (d) show H.264. The luminance component 16 × 16 intra prediction mode according to the H.264 / AVC standard and FIGS. 6A to 6I respectively illustrate the 4 × 4 intra prediction mode. As shown in FIGS. 4A to 4D, the 16 × 16 intra prediction mode includes a mode 0 vertical mode, a mode 1 horizontal mode, a mode 2 DC (direct current) mode, and a mode 3 plane mode. Includes 4 modes. Further, as shown in FIGS. 6A to 6I, the 4 × 4 intra prediction mode includes the mode 0 vertical mode, the mode 1 horizontal mode, the mode 2 DC mode, the mode 3 diagonal lower left mode, and the mode. There are a total of nine modes: 4 diagonal lower right mode, mode 5 vertical right mode, mode 6 vertical left mode, mode 7 horizontal upper mode, and mode 8 horizontal lower mode. Here, the pixels A to M are already encoded, and the pixel values of the adjacent pixels a to p shown in FIG. 6J are predicted using these pixel values.

  The predicted value of the target block is obtained by extrapolating adjacent encoded pixels along a plurality of directions having unique coefficients. For example, an operation for predictively encoding a target 4 × 4 block using the vertical prediction mode indicated by mode 0 in FIG. 6 will be described. First, the pixel value of the 4 × 4 current block is predicted based on the pixel values of the pixels A to D adjacent to the upper side of the 4 × 4 size current block. That is, the value of the pixel A is set to the pixel values of the four pixels a, e, i, m included in the first column of the 4 × 4 current block, and the value of the pixel B is set to the second column of the 4 × 4 current block. The pixel values of the four pixels b, f, j, and n included are changed, and the value of the pixel C is changed to the pixel values of the four pixels c, g, k, and o included in the third column of the 4 × 4 current block. The value of the pixel D is predicted to be the pixel value of the four pixels d, h, l, and p included in the fourth column of the 4 × 4 current block. In this way, the pixel values of the pixels a to p are all predicted using the pixels A to D. Next, the difference between the predicted 4 × 4 current block and the actual value of the pixels included in the original 4 × 4 current block is obtained, and this difference value is encoded.

  Such a fixed extrapolations method gives good results for continuous images with simple textures. However, this method cannot be applied to continuous images and contexts having complex edges within one macroblock.

  Here, Table 1 shows the result of simulation for the tendency that the intra mode is continuously selected. This table shows the buses, coastguards, footballs, foreman, mobiles, tamppets (standard) used as different types of video sources (sequences) tempete) shows a correlation in which an intra mode is selected between consecutive frames. In this table, “intra MB” indicates the number of macroblocks encoded in the intra prediction mode (modes 0 to 8), and “intra-intra MB” is in a correlation macroblock or a surrounding macroblock. The number of macroblocks encoded in the same intra prediction mode is shown.

  As shown in Table 1, the intra mode also has a very high tendency to continue between consecutive frames. Further, the intra-correlated correlation macroblock can be used to predict a reference pixel for encoding in the intra mode even in the current frame.

  Using such a prediction mode, the moving image data is compressed and encoded. In encoding, it is desirable to reduce the data size as much as possible. However, it is not easy to compress without impairing the quality of moving image data. Also, because of the high correlation between adjacent macroblocks, the macroblocks need to be encoded one by one in the order of macroblock numbers, that is, sequentially. Therefore, high-performance computing cannot be realized because macroblock-level parallel encoding processing cannot be used in a multi-core GPU.

  In general, it is considered that inter prediction is superior to intra prediction, and therefore intra prediction is employed when sufficient results cannot be obtained by inter prediction. However, in an image with little correlation, the number of bits may be insufficiently reduced in both inter coding and intra coding. In the test conducted by the present inventors, for example, in the case of a moving image of a football game, intra coding is adopted instead of inter coding even though there are similar parts between consecutive frames. I found out. For example, FIGS. 7A and 7C show continuous frame images in a moving image, and FIGS. 7B and 7D show encoding methods of the respective frame images. As shown in these figures, the present inventors have found that intra coding is used even though the lawn portion hardly changes between the previous and next frames.

  Furthermore, as shown in FIGS. 8A and 8B, when the encoding method is compared with the immediately preceding frame, the tendency of the prediction mode is unlikely to change between consecutive images, that is, the same prediction mode continues. I found out that it was done. From this, as shown in FIG. 9, when intra prediction is selected, it is highly likely that intra prediction is similarly adopted in the same position or in the vicinity thereof in the previous frame image. As a result of examining the rate at which intra prediction is continuously performed for various video sources, it is confirmed that intra prediction is continuous with a very high probability depending on the type of moving image, as shown in Table 1. I found it. From this, the idea that more efficient encoding can be performed by performing encoding using a result of intra encoding of the previous frame in a moving image in which intra prediction is easily selected. In other words, in addition to the correlation in the same image that has been performed by the conventional intra coding, the temporal correlation is also taken into consideration, so that further improvement in efficiency can be achieved. This embodiment is an independent prediction method as an intra mode for realizing parallel processing of block coding. Specifically, coding efficiency is improved by introducing a temporal prediction intra mode in addition to the existing intra mode. An algorithm for realizing such high-speed parallel coding and high coding efficiency will be described in detail below.

  In the present embodiment, it is preferable to use the original pixel instead of the encoded reference pixel. FIG. 10 shows the configuration of the reference pixel used here. As shown in this figure, it is assumed that pixels C0 to C15 constituting a macroblock are to be encoded, and adjacent pixels A to M have already been encoded. In general, intra mode reference data uses A to M encoded neighboring pixels to generate a reference block. On the other hand, in the present embodiment, the original pixel is used to generate a reference block so that RDO processing of one macroblock can be started without waiting for encoding of adjacent pixels. However, this causes an estimation error and impairs the encoding performance.

  FIG. 11 to FIG. 16 show simulation results comparing the performance when using such original pixels and when using encoded pixels. In these figures, FIG. 11 is the efficiency of the original pixel and the encoded image in the bus moving image, FIG. 12 is the efficiency of the original pixel and the encoded image in the coast guard moving image, and FIG. 13 is the football moving image. FIG. 14 shows the efficiency of the original pixel and the encoded image in the foreman moving image, FIG. 15 shows the efficiency of the original pixel and the encoded image in the mobile moving image, and FIG. Simulation results comparing the efficiency of the original pixel and the encoded image in the pet moving image are shown. In each graph, the horizontal axis represents the bit rate [kbit / s], and the vertical axis represents the peak signal-to-noise ratio (PSNR) [dB]. As shown in these drawings, the use of the original pixel causes a slight deterioration in encoding performance. However, it is a level that can be ignored from the viewpoint of the overall calculation amount, and is a level that can be sufficiently handled even by a conventional hardware configuration. On the other hand, by using original image data, the benefits obtained by enabling parallel encoding based on macroblocks are much greater.

  Further, the encoding of the first embodiment is performed on the bus, coast guard, football, foreman, mobile, and tamppet as the moving image sequence by using the original image data, and the bit reduction rate as well as the PSNR is compared with the comparative example 1. The comparison results are shown in Table 2. As shown in this figure, the bit rate was reduced in any moving image, and the average bit rate was reduced to 16.34%, which was an extremely excellent bit rate reduction as a single technique. Of course, further improvements can be obtained by combining other known encoding techniques.

(Procedure for temporal prediction intra mode)

Next, temporal prediction intra mode is demonstrated based on FIG.17 and FIG.18. According to the tests conducted by the present inventors, when the macroblock was encoded by intra prediction in the previous frame, it was found that the macroblock of the current frame has a deep correlation with this. Therefore, when correlated macroblocks are encoded in the intra mode, efficient 4D4 coding can be achieved by selecting one 4 × 4 block most similar to the current macroblock in the RDO processing of the current macroblock. can get. This technique is applied for intra prediction. Hereinafter, as shown in the schematic diagram of FIG. 9, when a macroblock in the current frame n is encoded, a procedure for using the encoded block existing at the corresponding position in the previous frame n−1 is used. This will be described based on the flowchart of FIG.
(Step S1: Determination of candidate reference block)

First, at step S1, candidate reference blocks are determined. Specifically, as shown in FIG. 18, it exists in the corresponding part in the previous frame n-1 that correlates with the current block to be encoded (shown in black in FIG. 18) in the current frame n. Attention is paid to 16 macroblocks of 4 × 4, and the intra prediction mode adopted most frequently among these blocks is selected as the most numerous mode (mode i). Next, all the blocks adopting this most numerous mode are extracted and set as candidate reference blocks. In the example of FIG. 18, assuming that the most numerous mode is mode 0, blocks X1, X2, and X3 encoded in this mode 0 are extracted as candidate reference blocks.
(Step S2: Selection of reference block)

  Next, in step S2, a reference block is selected from the candidate reference blocks. Here, all candidate reference blocks X1, X2, and X3 are compared with the current block to be encoded, and the block with the smallest difference is set as the reference block. The difference can be calculated by the following equation, for example.

In this way, one reference block that is most similar to the block to be encoded is selected.
(Step S3: Generation of second reference pixel)

  In step S3, second reference pixels A 'to M' are generated based on the reference block. The generation of the second reference pixel is performed by a method opposite to the pixel generation method used in each prediction mode 0 to 8 of the conventional 4 × 4 intra prediction. That is, as shown in FIG. 19 (j), the adjacent second reference pixels A ′ to M ′ are reversed based on the 4 × 4 pixel values a ′ to p ′ constituting the already generated reference block. Generate. The method of determining the pixel values of the second reference pixels A ′ to M ′ is performed based on nine additional modes (Ex modes) shown in FIGS. That is, here, nine sets of second reference pixels A ′ to M ′ are generated based on the pixel values of FIG.

  For example, in the additional mode 0 shown in FIG. 19A, a pixel value obtained by averaging the pixels a ′, e ′, i ′, m ′ located in the first column of the reference block is calculated as the second reference pixel A ′. Is done. Similarly, a pixel value obtained by averaging the pixels b ', f', j ', and n' located in the second column of the reference block is calculated as the second reference pixel B '. Similarly, the average values of the pixel values in each column are generated as second reference pixels C ′ and D ′, respectively.

  In addition mode 2 shown in FIG. 19C, the average of the pixel values of all the pixels a 'to p' of the reference block is input to the second reference pixels A 'to M', respectively. Table 3 shows a table showing the calculation method of each second reference pixel in each of these additional modes using mathematical values using the pixel values of FIG. In this way, the second reference pixels A ′ to M ′ are generated by using the pixel value of the reference block selected in step S <b> 2 by a method opposite to the existing intra reference pixel generation.

(Step S4: Generation of prediction pixel)

Then, based on the obtained second reference pixels A ′ to M ′, the same processing as in the conventional intra prediction mode is performed. That is, nine prediction modes 0 to 8 shown in FIGS. 6A to 6J are executed, and prediction pixels a to p are obtained based on the reference pixels A to M. In addition, second predicted pixels a to p are obtained based on the second reference pixels A ′ to M ′ generated in step S3. In this way, in addition to the nine existing intra prediction modes, the second prediction pixels a to p are newly calculated based on the second reference pixels A ′ to M ′ respectively calculated in the nine additional modes shown in FIG. Is generated.
(Step S5: Rate distortion optimization)

  Furthermore, rate distortion optimization is performed based on the obtained predicted pixels. That is, in addition to the conventional intra prediction modes 0 to 8, the cost is calculated based on the difference from the original pixel based on a total of 18 prediction pixels based on the additional modes 0 to 8, and RDO (Rate distortion optimization) ) Is performed. As a result, encoding with higher accuracy than conventional can be expected. With this method, although the amount of calculation increases theoretically as the additional modes 0 to 8 are added, the calculation processing of the second reference pixel and the calculation of the prediction pixel based on this are common in a considerable part. Therefore, the actual processing hardly changes. On the contrary, it can be said that the efficiency is good because the previous calculation result can be used appropriately. In addition, because it uses the same algorithm as the existing intra prediction mode, existing hardware can be used, and there is almost no cost penalty for installation, and it can be implemented at low cost on existing equipment. Benefits are gained.

(simulation result)
The temporal prediction intra mode according to this embodiment is referred to as H.264. The simulation was performed by mounting on JM (Non-Patent Document 5) used as the H.264 / AVC standard software. Here, “Bus”, “Coast Guard”, “Football”, “Foreman”, “Mobile”, and “Tamppet” were used as test moving images. The simulation conditions are as follows.
Codec: JM14.2
Resolution: CIF
Frame rate: 30Hz
Number of frames: 65
Profile: High Profile CABAC: ON
QP setting: 18-30
GOP size: 16
In these simulations, only the first frame is encoded as an I frame and the other frames are encoded as P frames.

  Here, the encoding performance was evaluated by an RD curve (Rate-Distortion Curve). 20 to 25 show RD curves as simulation results. In these figures, FIG. 20 is a bus moving image, FIG. 21 is a coast guard moving image, FIG. 22 is a football moving image, FIG. 23 is a foreman moving image, FIG. 24 is a mobile moving image, and FIG. The simulation results comparing the efficiency of the conventional encoding and the encoding according to the first embodiment in the pet moving image are shown. As is clear from these figures, the temporal prediction intra mode algorithm improves the image quality on average by 1 dB or more at the same bit rate as compared with the conventional JM. In the case of the same image quality, the bit rate was reduced by about 15%. This can be achieved by combining with other compression methods. This suggests the possibility of realizing a target reduction of about 50% with H.265 / AVC.

  As described above, H.264 mainly intended for implementation on GPUs. It was confirmed that the temporal prediction intra mode is effective as a new parallel processing algorithm of H.264 / AVC. That is, in order to realize complete macroblock level parallel processing of intra prediction, original pixel data is used in RDO processing instead of encoded pixel data. In addition, in order to compensate for the shortcomings of using spatial prediction and temporal prediction individually, we introduced a temporal prediction concept in the conventional intra mode. As a result, the simulation result shows that the encoding performance is improved by about 1 dB.

In the first embodiment, the additional modes 1 to 8 shown in FIGS. 19A to 19I are used as the reference pixel generation method. In each of the additional modes 1 to 8, the pixel values C0 to C15 shown in FIG. 10 are used to calculate the second reference pixels A ′ to M ′ as reference pixels according to the mathematical expressions shown in Table 3.
(Weighting)

  Here, it is considered that a more accurate reference pixel (third reference pixel) can be obtained by weighting the pixel value according to the distance between the target pixel and the reference pixel. Therefore, a moving image encoding method that employs a weighting mode (Wt-MODE) with weighting as an additional mode was performed as Example 2. Table 4 shows calculation methods of the third reference pixels A ″ to M ″ in the weighting modes 0 to 8 used in the second embodiment.

(location information)

  On the other hand, in the first embodiment, as shown in FIG. 18, when selecting a reference block from candidate candidate reference blocks, a macroblock at any position in the previous frame n-1 is selected as the reference block. The location information is not taken into account. For example, as shown in FIG. 26, the most selected additional mode in the macroblock of the previous frame n−1 is the mode x. Here, in the enlarged view of the reference macroblock, when blocks indicated by diagonal lines are encoded in mode x, these are set as candidate reference blocks. Then, as shown in FIG. 27, a block most similar to the current macroblock of the frame n to be encoded is searched from among the blocks Ref0 to Ref12 constituting the candidate reference block. For example, in the figure, when Ref0 is the most similar block, Ref0 is selected as a reference block and the position of this reference block is stored. By adding position information in this way, accurate image decoding can be performed.

  However, if position information is added, the number of bits increases accordingly, and the data size increases. For example, in the example of FIG. 28, since the reference block can take 16 positions from Ref0 to Ref15, 2 4, that is, 4 bits are added as position information to each block of the frame n to be encoded. FIG. 29 shows how the data size changes. In this figure, the hatched portion indicates the number of bits increased by the addition of position information. In particular, as shown in FIG. 30, when the position information of the reference block in the previous frame n-1 and the position information in the frame n to be encoded are stored so as to correspond to 1: 1, the position of the reference block Although the accuracy is improved, the number of generated bits increases. As a result, there is a problem that the final encoding efficiency is lowered.

  Therefore, in the second embodiment, the position information is efficiently added by collecting the position information of several macroblocks. Here, the macroblock is divided into sub-blocks, and for each of the divided sub-blocks, the top block is set as a key block, position information is set for this key block, and other than the key block The block is substituted with the position information of the key block. In the example shown in FIG. 31, the macroblock is divided into 2 × 2 sub-blocks. Further, the top 4 × 4 block of each sub-block is set as a key block. In the example of FIG. 31, the position information is set with the blocks key0, key1, key2, and key3 as key blocks. The blocks other than the key block in each sub block use the position information of the key block. With this method, since the position information can be reduced to ¼, an increase in the number of bits due to the addition of the position information can be suppressed.

  According to tests conducted by the inventors, it was confirmed that the reference blocks tend to exist in any sequence. FIG. 32 shows the result of calculating the same ratio of the position of the reference block and the other blocks on the basis of the top 4 × 4 block in units of sub-blocks for a plurality of standard moving image sources. As shown in this figure, it was confirmed that the position of the reference block can be used at 80% or more in any sequence. Therefore, it has been found that even if position information is added around the head block, encoding can be realized with practically no problem.

  FIG. 33 to FIG. 33 show simulation results comparing the efficiency of the encoding according to Example 2 and the encoding according to Comparative Example 1 described above with respect to a plurality of moving image sequences by the encoding method of Example 2 described above. 38. In these figures, FIG. 33 is a bus moving image, FIG. 34 is a coast guard moving image, FIG. 35 is a football moving image, FIG. 36 is a foreman moving image, FIG. 37 is a mobile moving image, and FIG. The simulation result which compared the efficiency by the encoding which concerns on Example 2 and the encoding which concerns on the comparative example 1 in the moving image of a pet is each shown. In each graph, the horizontal axis represents the bit rate [kbit / s], and the vertical axis represents PSNR [dB]. Further, PSNR difference (ΔPSNR [dB]) of the RD curve of each graph obtained by performing the encoding of Example 2 on the bus, football, foreman, tamper, mobile, and coast guard as a moving image sequence and Table 5 shows the bit reduction rate (Δbitrate [%]). From these results, the bit rate was reduced in any moving image, and an excellent bit rate reduction of 8.14% on average was realized.

  The moving image encoding apparatus, the moving image encoding method, the moving image encoding program, and the computer-readable recording medium of the present invention are useful for applications in which a moving image signal is compressed and recorded or transmitted. For example, it can be suitably used in devices that process high-quality moving images, such as mobile terminal devices, moving image capturing devices, moving image recording / reproducing devices, and their application fields.

DESCRIPTION OF SYMBOLS 100 ... Video coding apparatus 102 ... Motion estimation part 104 ... Inter prediction part 106 ... Intra prediction part 108 ... Transformer 110 ... Quantization part 114 ... Entropy encoding part 116 ... Inverse quantization part 118 ... Inverse transform part 119 ... RDO processing unit 120 ... filter 122 ... frame memory

Claims (15)

A moving image encoding method for encoding moving image data,
Obtaining moving image data;
In encoding an arbitrary macroblock constituting an arbitrary current image frame of the obtained moving image data for each subblock constituting the macroblock, the subframe in the previous image frame that has already been encoded is encoded. Extracting one or more candidate reference blocks highly relevant in the plurality of blocks from a plurality of blocks existing at a position corresponding to the block and its surroundings; and
Selecting from among the one or more extracted candidate reference blocks as a reference block a block that most closely approximates a sub-block to be encoded;
Generating a second reference pixel according to a plurality of pre-defined additional prediction modes based on a plurality of pixel values constituting the selected reference block;
Generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel;
Performing rate distortion optimization based on the second prediction pixel and selecting an intra prediction mode with the lowest rate distortion cost;
Encoding according to the selected intra prediction mode;
A moving picture encoding method comprising: The moving image encoding method according to claim 1,
In the step of generating the second prediction pixel, generating a prediction pixel acquired according to a predetermined intra prediction mode,
In the rate distortion optimization, an intra prediction mode with the lowest rate distortion cost is selected including a prediction pixel in addition to the second prediction pixel. The moving image encoding method according to claim 1 or 2,
The step of extracting the one or more candidate reference blocks selects the most commonly used intra prediction mode as the most numerous mode among the plurality of intra prediction modes employed in encoding the plurality of blocks, A moving picture coding method, wherein all blocks adopting the most number of modes are extracted as candidate reference blocks. A moving image encoding method according to any one of claims 1 to 3, comprising:
The step of generating the second reference pixel includes adding the pixels arranged in the vertical direction among the plurality of pixel values constituting the selected reference block and dividing the average value by the number of pixels into the second value A moving image encoding method, wherein the pixel value of a reference pixel is used. A moving image encoding method according to any one of claims 1 to 3, comprising:
The step of generating the second reference pixel includes adding the pixels arranged in the horizontal direction among the plurality of pixel values constituting the selected reference block and dividing the average value by the number of pixels into the second value A moving image encoding method, wherein the pixel value of a reference pixel is used. A moving image encoding method according to any one of claims 1 to 3, comprising:
In the step of generating the second reference pixel, an average value obtained by adding all the plurality of pixel values constituting the selected reference block and dividing by the number of all pixels is set as the pixel value of the second reference pixel. A video encoding method characterized by the above. A moving image encoding method according to any one of claims 1 to 6, comprising:
The plurality of additional prediction modes may be H.264. A moving picture coding method, which is an inverse operation of four intra prediction modes of 4 × 4 blocks of luminance components defined by the H.264 / AVC standard. A moving image encoding method according to any one of claims 1 to 6, comprising:
The plurality of additional prediction modes may be H.264. A moving picture encoding method, which is an inverse operation of four intra prediction modes of 16 × 16 blocks of luminance components defined by the H.264 / AVC standard. A moving image encoding method according to any one of claims 1 to 8, comprising:
A video encoding method characterized in that inter prediction mode is preferentially adopted over intra prediction mode in encoding, and the intra prediction mode is adopted when the inter prediction mode has a higher rate distortion cost than the intra prediction mode. . A moving image encoding method according to any one of claims 1 to 9, comprising:
A moving image encoding method, wherein original pixel data is used as image data to be encoded. A moving image encoding method according to any one of claims 1 to 10, comprising:
The step of generating the second reference pixel is a step of generating a third reference pixel in which a plurality of pixel values constituting the selected reference block are weighted according to the distance from the target pixel. A video encoding method characterized by the above. A moving image encoding method according to any one of claims 1 to 11, comprising:
In the step of selecting the reference block, for each of the sub-blocks, the first block is set as a key block, position information is set for the key block, and blocks other than the key block are key block positions. A moving picture coding method characterized by substituting information. Moving image input means for acquiring moving image data;
Compression means for compressing moving image data input by the moving image input means;
Quantization means for quantizing the compressed data compressed by the compression means;
A video encoding device comprising:
The compression means includes
In encoding an arbitrary macroblock constituting an arbitrary current image frame of the obtained moving image data for each subblock constituting the macroblock, the subframe in the previous image frame that has already been encoded is encoded. An extracting means for extracting one or more candidate reference blocks having high relevance in the plurality of blocks from a plurality of blocks existing at and around the position corresponding to the block;
A reference block selecting means for selecting, as a reference block, a block that most closely approximates a sub-block to be encoded from among one or more candidate reference blocks extracted by the extracting means;
Second reference pixel generation means for generating a second reference pixel according to a plurality of additional prediction modes defined in advance based on a plurality of pixel values constituting the reference block selected by the reference block selection means;
Second reference prediction pixel generation means for generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel;
Including
A moving picture code comprising: performing rate distortion optimization based on the second prediction pixel generated by the second reference prediction pixel generation means; and selecting an intra prediction mode having the lowest rate distortion cost. Device. A moving image encoding program for encoding moving image data, which is stored in a computer,
A function to acquire moving image data;
In encoding an arbitrary macroblock constituting an arbitrary current image frame of the obtained moving image data for each subblock constituting the macroblock, the subframe in the previous image frame that has already been encoded is encoded. A function of extracting one or more candidate reference blocks having high relevance in the plurality of blocks from a plurality of blocks existing at a position corresponding to the block and its surroundings;
A function of selecting, as a reference block, a block that most closely approximates a sub-block to be encoded from among the extracted one or more candidate reference blocks;
A function of generating a second reference pixel according to a plurality of pre-defined additional prediction modes based on a plurality of pixel values constituting the selected reference block;
A function for generating a second prediction pixel according to a predetermined intra prediction mode based on the second reference pixel;
A function of performing rate distortion optimization based on the second prediction pixel and selecting an intra prediction mode having the lowest rate distortion cost;
A function of performing encoding according to the selected intra prediction mode;
A moving picture encoding program characterized by realizing the above.   A computer-readable recording medium or a recorded device storing the program according to claim 14.