TW201143459A - Apparatus for encoding dynamic image and apparatus for decoding dynamic image - Google Patents

Abstract

This invention provides an apparatus for encoding dynamic image, in which a frequency information generating section 93 counts the occurrence frequency of a most appropriate encoding mode 7 and generates a frequency information 94, and a binarization table updating section 95 updates the corresponding relationship between a multi-values signal in the binarization table in a binarization table memory 105 and a binarization signal, base on the frequency information 94. A binarization section 92 converts a most appropriate encoding mode 7a in the multi-values signal into a binarization signal 103 by using the appropriately updated binarization table, and the binarization signal is encoded by a calculation encoding processing section 104.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding apparatus that divides a moving image into a predetermined area, encodes it in a unit of area, and a moving picture to be encoded in a predetermined area unit. A moving image decoding device that performs decoding. [Prior Art] The international standard video coding method such as MPEG (Moving Picture Experts Group) and "ITU-T Η, 26χ" is used to collect the luminance signal 16x16 pixels and the corresponding color difference signal 8χ8 pixels. The block data (called "macroblock") is a unit, and the method of compressing each frame of the image signal according to the motion compensation technique and the orthogonal transform 'transform coefficient quantization technique. The known motion compensation technique is to use the high degree of correlation between the video frames to reduce the redundancy in the direction of each money, and to use the frame that has been encoded in the past as the reference image. In the storage money body, the predetermined search range of the reference image towel is used to search for the current macro block and the smallest differential power block area that have become the target of the mobile compensation prediction. The offset of the spatial position of the search result block in the reference image is used as a technique for moving vector coding. In this conventional international standard image coding method, the area covered by the fixed block is easy to be covered by the large block of the large set of blocks, especially when the resolution of the image is high. Localized. Thus, the situation occurs when the peripheral macroblocks become the same coding mode and the same motion vector is assigned. In this case, the prediction efficiency will not increase' and 323005 4 201143459 *- will increase the total load of the encoded coding mode information and motion vector information, so the coding efficiency will be reduced as a whole for the encoder. . In order to solve such a problem, it is known that a device is formed by switching the size of a macroblock by the resolution of the artifact or the content (for example, refer to Patent Document 1). The moving image encoding device in Patent Document 1 is formed as The macro block size can be adaptively converted into a slice unit, a frame unit, a sequence unit, and the like in accordance with the image content, the resolution, the profile, and the like. The conventional international standard image coding method and the device described in the patent document i are related to the small locality by selecting the coding mode of the cut macroblock when there are different objects in the macroblock. Dynamic objects. However, the international standard image editing horse--, 1 is the formula, in addition to the macroblock partition must be edited: the information of the coding mode of the segmentation within the block is also in this regard, in the patent document j, the can be prepared by m In the larger block than the already determined macro block, the object is in a number of blocks in the frame, and the ratio of the coding mode in the tail of the selected macro is reduced. The way, due to the selection of the large giant block in the large county, the size of the block size in the macro block is large, and the same material &_; The relevant code quantity is not issued [previous technical literature] [patent literature] [Development of the International Public Gazette Xiayuan 34918 No. 323005 5 201143459 [Fa] The subject to be solved] To be determined = good! Question: In order to pass the content, it is very magical to evaluate the image correctly. In addition, the amount of processing related to the pre-processing of the month (3) is used to solve the problem of the image-free content as described above, and even if it is set in advance, the purpose is to provide a stone horse quantity that can suppress the total load such as the stone-horse mode, and In the size, the still encoded image encoding device and the moving image solution can efficiently compress the means for solving the problem. In the moving picture coding apparatus of the present invention, the predetermined portion selected according to the coding efficiency is selected by the coding mode value unit, and the coded root coding control unit of the variable length coding unit: number is selected: The binary signal converted by the Ministry is in the middle of the division, the system is in the ranks, and the coded bits are listed as exhausted: the bat stone update table is rotated, according to the coding mode. And the binary frequency 'to slant more of the aforementioned binarization table = code: the choice relationship of the department. The moving picture decoding device corresponding to the time-and-value signal of the present day and the month is subjected to arithmetic-depletion by the variable-length decoding unit to generate a binary signal by the arithmetic decoding solution. The modulo u 'reverse binary value 323 〇〇 5 201143459 is a binary signal that is generated by the arithmetic decoding processing unit using a binarization table that specifies a correspondence relationship between the binary signal indicating the coding mode and the multi-value signal. The coding mode indicated to be converted into a multi-value signal. [Effect of the Invention] According to the present invention, since the binarization table updated by the selection frequency of the encoding control unit according to the encoding mode is used, the multi-value signal indicating the encoding mode is converted into a binary signal for arithmetic coding, and more Industrialization to the bit stream. Therefore, it is possible to provide an image-capturing device that can efficiently suppress the coding mode, such as the coding mode, and the like, even in the preset macroblock size. Like a decoding device. [Embodiment] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. (First Embodiment) In the first embodiment, motion compensation prediction is performed between adjacent frames with respect to each frame image using video, and orthogonal compression-quantization compression is applied to the obtained prediction difference signal. After the processing, a moving image encoding device that performs variable length decoding to generate a bit stream and a moving image decoding device that decodes the bit stream will be described. Fig. 1 is a block diagram showing the configuration of a moving image encoding device according to a first embodiment of the present invention. The moving image encoding device shown in Fig. 1 includes: a block dividing unit 2 that divides each frame image of the input image signal 1 into a block image of a plurality of blocks of the macro block size 4, According to the coding mode 7, the macro/sub-block image divided into one or more sub-blocks is output 7 323005 201143459 5; the in-plane prediction unit 8 is input when the macro/sub-block image 5 is input. The macro/sub-block image 5 is intra-frame predicted using the imaging signal of the intra-frame prediction memory 28 to generate a prediction artifact 11; the motion compensation prediction unit 9 is connected to the macro/sub-region. In the case of the block image 5, the reference image 15 of the motion compensation prediction frame memory 14 is used, and the macro/subblock image 5 is subjected to motion compensation prediction to generate a prediction artifact 17; the switching unit 6 is in response to the coding mode 7 The macro/subblock image 5 is input to either the in-plane prediction unit 8 or the motion compensation prediction unit 9; the subtraction unit 12 is a macro/subblock image 5 outputted by the block division 4 2 The prediction images u and 17 output from either one of the in-plane prediction unit 8 or the motion compensation prediction unit 9 are subtracted to generate a prediction difference. The signal-to-transformation-quantization unit 19 performs a text conversion and quantization process on the prediction difference signal 13 to generate a compressed data u. The variable-length coding unit f entropy-codes the compressed data 21 into a bit stream 30. The inverse transform unit 22 inverse-quantizes and inversely converts the compressed data 2 to the local decoded differential signal 24; the addition unit 25 adds the intra-frame prediction unit 8 or the motion compensation to the any-H22 of the inverse amount. The prediction unit 9 generates a local gamma image for the prediction images 11 and 17 generated by the searcher, and generates a partial solution image (si), and the Zb, the loop filter unit 27 processes the local decoded image = image signal 26 Filtering is performed to store the local decoding image 29 and the shifting frame memory encoding control unit 3, which is necessary for outputting the block size 4 and the encoding mode 7 = 1 (major best prediction) The parameter 1〇, the code mode is, the prediction parameter 10, the number and the 18a 1 contraction parameter 2 (), the optimal compression parameter 32350 8 201143459, 20a). Hereinafter, the details of the macro block size 4 and the coding model will be described in detail later. /, the encoding control unit 3 specifies the macro block size 4 of each frame image of the input image signal 1 for the block dividing unit 2, and indicates the coping image according to each macro block of the encoding object All encoding modes selected by type 7. Further, the hummer control unit 3 may select a pre-coding mode from among the coding mode groups, but the coding mode group is arbitrary, and is set to be selected by, for example, a group of the 2A picture or the 2B picture shown below. Encoding mode. Fig. 2A is a diagram showing an example of a coding mode of a p Gfedictive) image in which prediction in the time direction is performed. In Fig. 2A, this mode 〇 to 2 is a mode (inter) in which a macroblock (MxL pixel block) is encoded by inter-frame prediction. Mb_mode 0 is a mode in which one motion vector is allocated to the entire macroblock, and mb_mode 1 and 2 are horizontally or vertically equal-sized macroblocks, and different mobile vectors are allocated in the divided sub-blocks. Mode. Mb_jnode 3 is a mode in which a macroblock is divided into four groups, and a different coding mode (subjnb-mode) is assigned to each divided sub-block. Sub-mbjnode 0 to 4 is to allocate rab~m〇de 3 in the coding mode of the macroblock, and allocate each sub-block (m Μ pixel block) obtained by dividing the macroblock into 4 The coding mode, and sub_jnb_jn〇de 0 is an intra (intra) coding of the sub-block by intra-picture prediction. In addition, for the mode (inter) coded by inter-frame prediction, sub_mb_mode 1 assigns a pattern of motion vectors to all the wild sub-blocks, and subjnbjnode 2, 3 are horizontal or vertical molecular blocks, respectively. And in each of the divided sub-areas 9 323005 201143459, the blocks are respectively assigned different modes of moving direction, sub_mb_m〇 (ie 4 is 4 divided sub-blocks), and each sub-block is divided into different sub-blocks. The mode of the motion vector. In addition, FIG. 2B is another example of the coding mode for displaying the p-picture for predictive coding in the time direction. In FIG. 2B, mbjnode 〇 to 6 is a macro set by inter-frame prediction. The block (MxL pixel block) is edited and the mode (inter). mbjnode 0 is a mode for assigning one motion vector to the entire macro block' mb-mode 1 to 6 are horizontal, vertical or diagonal directions respectively. The segmentation macroblock' assigns a pattern of different motion vectors to each of the divided sub-blocks. mb_mode 7 divides the macroblock into 4 segments, and assigns different brewing modes to the divided sub-blocks. Model of (sub_mb_mode) Sub-mbjnode 0 to 8 is a compilation mode in which each sub-block pixel block obtained by dividing the macroblock block is selected in the encoding mode of the macroblock block, subjnl3_m〇de 〇 An intra (intra) coded sub-block by intra-frame prediction. In addition to the inter-frame prediction coding mode (inter), sub-mb_m〇de 1 is a mode for assigning one motion vector to the sub-blocks, ___« 2 to 7 / knife is horizontal, The sub-blocks are divided vertically or diagonally, and the different sub-blocks are respectively assigned different modes of motion vectors, and mode 8 is to divide the sub-blocks into 4, and the different moving vectors in the divided Mode. Referring to the block dividing unit 2, each frame image of the input image signal 1 that is rotated in the moving image coding device is divided into macros of the size of the block 323005 10 201143459 by the coding control unit 3. Block image. Further, the block dividing unit 2 includes a mode in which the encoding mode 7 specified by the encoding control unit 3 is different in the encoding mode of the sub-blocks obtained by dividing the macroblocks (sub_mb_mode 1 to 4 in FIG. 2A). Or in the case of sub_mb_mode 1 to 8) of Fig. 2B, the macroblock image is divided into sub-block images shown in the coding mode 7. Therefore, the block image outputted by the block dividing unit 2 becomes either a macro block image or a sub-block image in response to the encoding mode 7. Hereinafter, this block image is referred to as a macro/subblock image 5. In addition, when the horizontal or vertical size of each frame of the input image signal 1 is not an integer multiple of each horizontal size or vertical size of the macro block size 4, each frame of the input image signal 1 is generated in the horizontal direction. Or expand the frame of the pixel in the vertical direction (expanded frame) until the frame size becomes an integer multiple of the size of the macro block. In the method of generating pixels in the extended area, for example, when the pixels are expanded in the vertical direction, the pixels at the lower end of the original frame are repeatedly filled, or have fixed pixel values (gray, black, white, etc.). The pixels are filled and the like. Similarly, in the case where the pixels are expanded in the horizontal direction, there are methods in which pixels on the right end of the original frame are repeatedly filled, or pixels having fixed pixel values (grey, black, white, etc.) are filled. The expansion frame of the frame size of the input image signal 1 is an integer multiple of the macro block size, and can be input to the block division unit 2 instead of the frame image input to the image signal 1. In addition, the frame size (horizontal size and vertical size) of each of the macroblock block size 4 and the input image signal 1 is multiplexed in the order unit or image unit formed by the image above the 1 frame. Turn into a bit stream, so it can be 11 323005 201143459

In addition, the value of the macro block size may not be directly multi-streamed, but may be specified by a profile. In this case, the identification information for the profile of the Chuanchuan staff is multiplexed into a bit stream. Used to identify units in sequence

(Hereinafter referred to as the inter-frame prediction mode), the macro/subblock image 5 is input to the motion compensation side portion 9. The in-plane prediction unit 8 is for the macro of the input person, and the sub-block image 5' is a macro block of the coding target specified by the macro block size 4 or a sub-block specified by the coding mode 7. , for intra-frame prediction. Further, the in-plane prediction unit 8 uses the image in the frame stored in the in-plane prediction memory 28 for all the picture Φ intra prediction modes included in the prediction parameter 1 指示 instructed by the coding control unit 3. Signals, respectively, to generate prediction artifacts u. Here, the prediction parameter 1〇 is explained in detail. When the coding mode 7 is the intra-frame prediction mode, the coding control unit 3 specifies the intra-plane prediction mode as the prediction parameter 1 对应 corresponding to the coding mode 7. The intra-plane prediction mode has, for example, forming a macroblock or a sub-block into a 4 pixel block unit, and generating a prediction image using pixels around the unit block of the image signal in the intra-screen prediction memory 28. a mode in which a macroblock or a sub-block is formed into an 8x8 pixel block unit, and a mode of predicting an image is generated using pixels around a unit block of the image signal of 12 323005 201143459 in the intra-screen prediction memory 28; The macroblock or the sub-block is formed into a 16x16 pixel block unit, and the pixel is generated by using the pixels around the unit block of the image signal in the memory 28; the mode of predicting the image is reduced; A pattern obtained by generating a predicted image in a block or a sub-block. The motion compensation prediction unit 9 specifies the reference image 15 for generating the prediction artifact from the reference image data of one frame or more stored in the motion compensation prediction frame memory 14, and uses the reference image 15 and The macro/subblock image 5 is subjected to motion compensation prediction in the coding mode 7 instructed by the coding control unit 3, and the prediction parameter 18 and the prediction artifact 17 are generated. Here, the prediction parameter 18 will be described in detail. When the coding mode 7 is the inter-frame prediction mode, the motion compensation prediction unit 9 obtains the motion vector, the identification number of the reference image (refer to the imaging index) indicated by each motion vector, and the like as the corresponding coding mode. Prediction parameter 18 of 7. The generation method of the prediction parameter 18 will be described in detail later. The subtraction unit 12 subtracts any one of the prediction artifact 11 or the prediction artifact 17 from the macro/subblock image 5 to obtain the prediction difference signal 13. Further, the prediction difference signal 13 is generated for each of the prediction artifacts 11 generated by the in-plane prediction unit 8 in response to all the in-plane prediction modes specified by the prediction parameter 10. The prediction difference signal 13 generated in response to all the intra-screen prediction modes specified by the prediction parameter 10 is evaluated by the encoding control unit 3 to determine the optimum prediction parameter l〇a including the optimum intra-picture prediction mode. For the evaluation method, for example, the predicted difference signal 13 is transformed by 13 323005 201143459 and the compressed data 21 obtained by the selection is used to calculate the in-screen cost, which is described later, and the flat code cost h is minimized. . The horse-horse control unit 3 predicts in-plane prediction: for all modes included in the coding mode 7, the evaluation == 13 ' according to the evaluation result, the coding efficiency is best obtained by the coding mode 7 The V-Decision Department 3' is composed of prediction parameters 1〇, 18 and 2:. The second and second codes are controlled by the optimal prediction parameters lQa, 18a and the most = parameter 2Ga in the optimum coding mode 7a. The order of each decision is described later. money

Heart: and the above, in the case of the prediction mode in the frame, the prediction m and the optimal prediction parameter include the in-plane and J-direction, and the mode in the inter-frame prediction mode. In addition, the measurement parameter 18a includes a motion vector, a number 18, and an optimal pre-identification number (refer to the imaging indicator). Reference image of the & in addition, compression parameter 20 and optimal compression parameters 2 〇 block size, quantization step size, and so on.匕3 has the result of transforming the order determined by this, the optimal encoding of the macroblock or sub-block of the encoding control, the optimal encoding parameters 10a and l8a, the optimal compression parameter 2〇a output to the second-prediction prediction parameter, and the encoding The control unit 3 outputs the compression parameter S3 code unit to the conversion-quantization unit 19 and the inverse quantization and inverse=the reduced parameter conversion-quantization unit 19, and the plurality of prediction difference signals 133 generated from the corresponding seeding mode. All the control units 3 mosquitoes (4) her _% delete the 323005 14 201143459 18a and the predicted image η and 17 corresponding to the predicted differential signal 13 (hereinafter referred to as the best predictive differential signal (4), and according to the editor The transform block size of the optimum compression parameter 2〇a determined by the unit 3 is subjected to DGT (Discrete Cosine Transform) processing or the like for the optimum_differential signal 13a to calculate a transform coefficient, and is instructed by the encoding control unit 3 The quantization step size of the optimum compression parameter 20a is used to output the compressed data 21 belonging to the quantized transform coefficient to the inverse transform unit 22 and the variable length coding unit 23. The inverse-inverse transform unit 22 , using the best compression The number 2 = the compressed data 21 input from the unit 19 is inversely quantized by the inverse transform processing such as the inverse DCT, and the predicted difference signal 13a = the real decoded prediction difference signal 2 4 is generated and output to the addition unit 25. p is like the second method: 25' is to add the local decoded prediction difference signal 24 to the prediction image or the U or the prediction image 17 to generate a local solution = loyalty to this partial solution to the full German image such as the image 2 2 exists in the whole ^ The image is output to the loop filter, the waver 27, and μ is the image signal for the intra-frame prediction. ...,,, the image signal 26 is the loop filter 27, and the pair is subjected to the predetermined image wave 26 for predetermined chopping. The local decoding image processed by the input is stored in the mobile complement (4), the target ^, and the local decoding 29 after the wave processing becomes the motion compensation; the measurement is performed in the chopping sound of the 27th, and the image is tested. 15. The partial solution of the loop filter block (4) input == line set area 323005 201143459 The variable length coding unit 23 outputs the compressed data 21 outputted by the transform-quantization unit 19 and output by the code control unit 3. Optimal coding mode 7 a, best prediction parameters 10a and 18a, and optimal pressure The parameter 20a performs entropy coding, and generates a bit stream 30 displaying the result of the encoding. Further, the optimal prediction parameters 10a and 18a and the optimal compression parameter 20a are in units of coding modes corresponding to the optimal coding mode 7a. In the moving image encoding device of the first embodiment, the motion compensation prediction unit 9 and the conversion-quantization unit 19 are operated in conjunction with the encoding control unit 3, and it is determined that the optimum is obtained. The coding mode, prediction parameters, compression parameters (i.e., optimal coding mode 7a, optimal prediction parameters 10a and 18a, optimal compression parameters 20a) of coding efficiency. Here, the order of determining the coding mode, the prediction parameter, and the compression parameter for obtaining the optimum coding efficiency by the coding control unit 3 is explained by the order of 1. prediction parameter, 2. compression parameter, and 3. coding mode. . 1. Determining the order of the prediction parameters Here, when the coding mode 7 is the inter-frame prediction mode, the motion vector including the inter-frame prediction and the identification number of the reference image indicated by each motion vector are determined (refer to the image). The order of the prediction parameters 18 such as the indicator). The motion compensation prediction unit 9 is associated with the coding control unit 3, and all the coding modes 7 instructed by the coding control unit 3 to the motion compensation prediction unit 9 (for example, the coding mode group shown in FIG. 2A or FIG. 2B). Determine the prediction parameter 18 separately. Hereinafter, the detailed order will be described. Fig. 3 is a block diagram showing the internal configuration of the motion compensation prediction unit 9. The motion compensation prediction unit 9 shown in Fig. 3 includes a motion compensation area 16 323005 201143459 (four) unit 40, a motion detecting unit 42, and an inner illustrator generating unit 43. Further, the input data includes an encoding mode 7 input by the encoding control unit 3, a macro/subblock image 5 input by the switching unit 6, and input from the motion compensation prediction frame memory 14. Reference portrait 15. In the coding mode 7 instructed by the coding control unit 3, the macro/sub-block image 5 input by the switching unit 6 is divided into blocks forming a unit of motion compensation, and this is determined by the coding mode 7 indicated by the coding control unit 3. The motion compensation block block 41 is output to the motion detecting portion 42. The internal illustrator image generating unit 43' is a reference image data towel finger (four) of one frame or more stored by the motion compensation prediction frame memory 14 to generate a reference image 15 of the predicted image, and the motion detecting unit 42 is located at A motion vector 44 is detected within a predetermined motion search range on the designated reference image 15. Further, the detection of the motion vector is performed by the motion vector of the virtual sample precision as in the MPEG_4 AVC standard or the like. This detection method is based on the pixel information (called an integer image) of the reference image, and creates a virtual_count (pixel) between integer pixels = interpolation, and uses the virtual 'pixel' as a prediction. Image to use. In the MPEG-4 AVC specification, ΓνΓ=18 is used as a virtual sample of readness. In addition, in (4)-4, the horizontal yd 'j/2 pixel precision imaginary can be used in the vertical direction or 6 integer pixels. Six taps ". _ ferrites generate 1/2 I / contiguous. 1/4 pixel precision virtual samples are obtained by using averaging or integer pixels. (4) Motion compensated prediction of the first embodiment The portion "' is also generated by the interpolation 323005 201143459. The artifact generation unit 43 generates the predicted image milk of the virtual pixel corresponding to the accuracy of the finger 44 indicated by the moving sample 9. : Example of moving vector detection order of pseudo pixel precision. ♦ Virtual (moving vector detection sequence I) The interpolated image generating unit 43 generates a photo book 45 of the == motion vector with respect to the search axis of the shift block image 41. Pre-image = (predicted image (7) wheel (four) subtraction unit 12, which is subtracted by the subtraction unit i2 ==:41 (major, sub-block image 5), becomes the pre-control unit 3 The prediction efficiency of the prediction difference signal 13 and the integer! (prediction parameter 18): The evaluation of the prediction efficiency is determined by, for example, the following formula (m-calculated prediction cost L, and = mobile search #_ Accurate movement vector 44. Donation (1)

Ji = Di + λ Ri 2 is set to use Dl, Rl as the evaluation value. L is the prediction difference = the absolute value in the macroblock of the signal or in the sub-block and (10)) 4 is the motion vector and the estimated code quantity λ· of the reference image_ job code indicated by the (four) vector is a positive number. When the current value of the mobile vector is taken, the code quantity of the motion vector uses the value of the nearby motion vector to predict the value of the vector of each mode of the 2A picture or the 2b picture, and the predicted difference value is borrowed according to the probability distribution. It is determined by the smoke code, or the input 彳τ is equivalent to the estimation of the code amount. Fig. 4 is a view for explaining a method of determining a predicted value (hereinafter referred to as a predictive vector) of a motion vector 323005 18 201143459 of each coding mode 7 shown in Fig. 2B. In the rectangular block of mb_mode 0, sub_mb_mode 1 and the like in FIG. 4, the respective encoded motion vectors MVa, MVb located at the left horizontal (position A), upper (position B), and upper right (position C) are used. MVc, the prediction vector PMV of the rectangular block is calculated by the following equation (2). The medianO is a function corresponding to the median filtering process and outputs the median of the motion vectors Mva, MVb, and MVc. PMV = median (MVa, MVb, MVc) (2) On the other hand 'on the diagonal block with diagonal shape! Nb_mode 1, sub_mb_mode 2, mb_mode 2, sub-mb_mode 3, mb_mode 3, sub-mb_mode 4, mb_mode 4, sub_mb "in the case of node 5" is formed to be applicable to the same rectangular block Since the processing is performed, the positions of the positions A, B, and C at which the median value is obtained can be changed by coping with the angular shape. Thereby, the method itself for calculating the prediction vector PMV can be changed without changing the shape of each motion vector allocation region, and the cost of the evaluation value 匕 can be suppressed to a small value (moving vector detection sequence II). The part 43 generates a predicted book image 45 for five or more "/2 pixel precision motion vectors 44 located around the motion vector of the integer pixel precision determined by the "moving vector detection order I". In the same manner, the image 17) is generated by the subtraction unit 12 by the subtraction unit 12, and the subtraction unit 12 is subtracted from the motion compensation sub-block image 5 to obtain the 13- and 2-pixel " 3 The prediction difference signal 13 and the _recording degree __44 (the evaluation of the efficiency of the _parameter (8) lai 323005 19 201143459' is determined by the motion vector of 1 precision around the motion vector of the integer pixel precision, so that the prediction cost l becomes 竑a motion vector 44 having a small 1/2 pixel precision. (Moving Vector Detection Sequence 111) The moving unit 3 and the motion compensation prediction unit 9 also have the same 1/4 pixel accuracy as described above. One or more 1/4 pixels around the motion vector of the motion vector detection order are predicted to be (moved vector detection order Iv). In the same manner, the coding control unit 3 and the motion compensation prediction unit 9 are combined. The motion vector of the virtual pixel precision is measured until the predetermined accuracy is obtained. In the present embodiment, the shift of the virtual pixel accuracy is detected (four) until the predetermined accuracy is obtained, but for example, the threshold value relative to . The prediction cost J1 is set to be smaller than the predetermined threshold, and the movement frame L is stopped before the predetermined accuracy is reached. The amount may also be referred to by reference to the frame size. The pixels outside the frame must be generated. In terms of the generation method of the image, there is a pixel filled at the end of the screen. In addition, the size of the frame of each frame in which the image signal 1 is rotated is not huge = block When the integer part of the size is substituted, the frame of the input image 1 is replaced, and when the frame is expanded, the size is 323005 20 201143459 which is an integral multiple of the size of the macro block (expanded frame) The size is the frame size of the reference frame. On the other hand, in the case of not referring to the local decoding portion of the extended region, only the local decoded portion relative to the original frame is referred to as a pixel in the frame, The frame size of the frame is the frame size of the original input image signal. Thus, the motion compensation prediction unit 9 divides the macro/sub-block image into the coding mode 7 and becomes the unit of motion compensation. The motion compensation area block image 41 of the block unit outputs the motion vector of the virtual pixel precision of each predetermined precision determined and the identification number of the reference image indicated by the motion vector as the prediction parameter 18. Further, the motion compensation prediction The portion 9 outputs the prediction artifact 45 (predicted image 17) generated by the prediction parameter 18 to the subtraction unit 12, and subtracts the macro/subblock image 5 from the subtraction unit 12 to obtain a prediction difference signal. 13. The prediction difference signal 13 output from the subtraction unit 12 is output to the transform-quantization unit 19. 2. Determination procedure of compression parameters Here, the prediction difference signal 13 generated based on the prediction parameter 18 determined for each coding mode 7 in the above-mentioned "1. Determination order of prediction parameters" is used for conversion-quantization processing. The order in which the compression parameter 20 (transform block size) is determined. Fig. 5 is a view showing an adaptation example of the conversion block size in accordance with the coding mode 7 shown in Fig. 2B. In Fig. 5, for the MxL pixel block, a 32x32 pixel block is taken as an example. When the mode specified by the coding mode 7 is mb_mode 0 to 6, the transform block size can adaptively select either of the 16x 16 or 8x8 pixels. When the coding mode 7 is mb_mode 7, 21 323005 201143459 (4) (4) The size can be adaptively selected from 8x8 or 4x4 pixels by the wrist 6 pixel sub-block obtained by dividing the macroblock by -4. In addition, the transform block size that can be selected in the parent-specific coding mode can be defined by the size of the rectangular block below the sub-block size that is equally divided by the coding mode. Fig. 6 is a diagram showing the mode of the coding mode according to the size of the block shown in the first figure, and the mode of the coding mode " For the aforementioned mb-mode 〇, 5, 6, in terms of the size of the selected transform block, except _6; word = the sub-unit of the mobile compensation unit" Bu and the small. In the case of: the second The transform block is adaptively selected in large and medium. It is adaptively selected in mb X8 and 32x32 pixels = ship 6, 8x8, 32xlfi German! ' 崎6's situation can be selected by the map and the medium In addition, although omitting 2 can be adaptively selected by two image materials, it is selected in _ prime, and == two _, for the rectangle, the selection is adapted. -,] by 8x8, 4x4 pixels, the mode == system will correspond to the fifth The map and the illustrated block size group illustrated in Fig. 6 are used as the compression parameters 2〇. In addition, in the examples of Fig. 5 and Fig. 6, the field is in the mouth of the wide code mode 7 and the premise Mm is large. The unit or sub-block unit of the cluster block can be adaptively selected, but the same is 323005 22 201143459: the sub-block obtained by cutting the macro block The coding mode (mode! to 8, etc.): the _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The conversion block of the coding mode 7 should be adaptively selected. The open conversion-quantization unit 19 is associated with the coding control unit 3, and is in the sub-block unit specified by the size 4 or in the coding mode. 7 into one: cut the sub-block unit of the macro block unit, and determine the optimal transform block size from the transform block size. The detailed sequence is explained below. Figure 7 shows the transform-quantization unit. The block diagram of the internal structure of Fig. 7. The transform-quantization unit 19 shown in Fig. 7 includes a transform block size division unit 50, a conversion unit 52, and a quantization unit 54. Further, as for the input data, there is an encoding control unit. The input compression parameter 20 (the conversion block size and the quantization size, etc.) and the prediction difference signal 输入 input by the coding control unit 3. The conversion block size division unit 50 is an object belonging to the object that determines the size of the conversion block. Prediction of a macroblock or subblock The sub-symbol 13 is converted into a block corresponding to the transform block size of the compression parameter 20, and is output as the transform target block 51 to the transform unit 52. Further 'in the compression parameter 20 for one macro block or sub-block selection In the case where a plurality of blocks are changed, the conversion block 51 of each transform block size is sequentially output to the conversion unit 52. The conversion unit 52 is for the input conversion target block 51, The transform coefficient 53 is output to the quantization unit 54 in accordance with 23 323005 201143459 DCT, integer transform such as integer transform DCT transform coefficient, and Hadamard transform. The quantization unit 54 quantizes the input transform key 53 according to the quantization step size of the compression parameter 2〇 indicated by the editing control unit 3, and outputs the compressed data 21 of the transformed transform coefficient to the inverse quantization_reversal. ^ and the encoding control unit 3. Further, the conversion unit 52 and the quantization unit 54 perform the above-described processing for all of the conversion block sizes in the compression parameter (4)^========================================================================= And output each compressed data 21. The compressed data 21 outputted by the quantizing unit 54 is input to 'the encoding efficiency for the (four) block size of the compression parameter 2Q, and is used in relation to all the voltages selected for the encoding mode 7 The shrinkage material 21' obtained by changing the block size is calculated by, for example, the following formula (3), and the size of the transform block whose code cost L is the smallest is selected. J2==D2+ AR2 (3) At #^, use D2 and R2 as evaluation values. In the case of D2, the compressed data 21 obtained by using the block size is input to the inverse quantization 'reverse 栌, 卩22, and the compressed data 21 is subjected to inverse quantization _ inverse transform processing to obtain the local decoded prediction difference. The image obtained by the 24th person prediction image 17 is obtained; the distortion between the code image signal 26 and the macro/subblock image 5 is obtained. For R2, 'the amount of code used is the size of the transform block. 24 323005 201143459 The compressed data obtained 2 and the compressed data 21 are related to the encoding mode 7 and the prediction parameters 10, 18, the scientific variable length encoding unit 23 is actually encoded. The amount of code obtained (or the estimated code size). The coding control unit 3 determines the optimum coding mode 7 in accordance with the "3. coding mode determination order" which will be described later, and the conversion block size of the determined optimal coding mode 7a is included in the optimum compression parameter coffee. And output to the variable length coding unit 23. The variable length coding unit 23 multiplexes the optimum compression parameter 20a into a bit stream 3〇 after entropy coding. Here, the 'transform block size' is selected by the transform block size group (exemplified in FIG. 5 and FIG. 6) defined in advance according to the best coding mode 7a of the macroblock or sub-block, so that each _ The _transform block size group assigns identification information such as ID to the transform block size included in the group, and entropy encodes the identification information as information for transforming the block size, and then multiplexes it into a bit stream 70G 3G Just fine. In this case, the identification information of the block size group is also decoded. ^Transform block size group includes changes: 奂 area, when the size is one, since the decoding device size can be automatically determined by the set on the decoding device side, it is not necessary to identify the block size. Information multiplexing into a bit stream 3 〇. ... 3. Determining the encoding mode According to the above 1. Predicting the order of the parameters, the order is determined, and the encoding is controlled (4). Determination of the H-contraction parameter All the coding modes 7 of the prediction parameters 1G and 18^3^, respectively, when the respective coding ports are used, the coding control unit 3 obtains the prediction parameter 1〇, which is obtained from the reduction parameter 2〇, 18 and the pressure pre-/difference signal 13 into a conversion-quantization 323005 25 201143459 obtained compression data 21, from the above equation (3) code mode 7, and select its coding mode 7 ,, horse cost (four) small coding mode 7a. In addition, it is also possible to determine the best in the 2nd map or the (skip mode) ^ ^ ^ ― 曰 在 . . . . . . . . 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接As a partial two = i industrialization into a bit * * M material _ 丝 丝 丝 丝 并 并 并 并 并 并 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝 丝In the same order, the second order of the 解 戍 戍 使用 使用 使用 使用 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 邻接 解 解 解 解 解 解 解 解 解 解 解 解 解 解 解 解The size of the frame is not the integer of the macro block h size 聋q input expansion frame; input frame 1 block 0 block, Θ n , , can be & The block becomes sub-tail I, and the skip mode (10) mode is selected to suppress the amount of code consumed by the extended area. The 疋 coding and coding control unit 3 increases the efficiency of the Μ 序 序 、 「 2 2 2 2 2 2 上 上 上 上 上 上 上 上 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 ; ; ; ; The optimum encoding mode 7a output: predicting the second F sentence prediction parameter 1 〇 18 The compression parameter 20 corresponding to the optimal encoding mode 7a is also subjected to the variable length encoding unit 23 as ^= 323005 26 201143459 parameter 20a. The variable length coding unit 23 entropy decodes the optimum coding mode 7a, the optimum prediction parameters 10a and 18a, and the optimum compression parameters 20a, and multiplexes them into the bit stream 30. Further, the optimum predicted difference signal 13a obtained from the predicted artifacts 11 and 17 based on the determined optimum coding mode 7a and the optimum prediction parameters 10a, 18a and the optimum compression parameters 20a is as described above. The quantization unit 19 converts and quantizes the compressed data 21. This compressed data 21 is entropy encoded in the variable length coding unit 23, and multiplexed into the bit stream 30. Further, the compressed data 21 is subjected to the inverse quantization-inverse conversion unit 22 and the addition unit 25 to become the local decoded artifact signal 26, and is input to the loop filter unit 27. Next, a moving image decoding device according to the first embodiment will be described. Figure 8 is a block diagram showing the configuration of a moving image decoding device according to the first embodiment of the present invention. The moving picture decoding apparatus shown in Fig. 8 includes a variable length decoding unit 61 that entropy decodes the optimum encoding mode 62 in a macroblock unit by the bit stream 60, and performs the best according to the decoding. The macroblock or sub-block unit divided by the coding mode 62 entropy decodes the optimal prediction parameter 63, the compressed data 64, and the optimal compression parameter 65; the in-plane prediction unit 69 is based on the optimal prediction parameter 63. At the time of input, the prediction image 71 is generated using the in-plane prediction mode included in the optimal prediction parameter 63 and the decoded image 74a stored in the in-plane prediction memory 77. The motion compensation prediction unit 70 is based on the optimal prediction. When the parameter 63 is input, the motion vector included in the optimal prediction parameter 63 and the reference in the motion compensation prediction frame memory 75 specified by the reference imaging index included in the optimal prediction parameter 63 are used. The image 76 performs motion compensation prediction and generates a prediction 昼 27 323005 201143459 Image 72, the switching unit 68 is the best prediction parameter 63 rounded by the decoded optimal code 桓 variable length decoding unit 61, and will Measuring unit 69 or Any one of the motion compensation prediction units 70; the intra-screen pre-stage 66 uses the optimal compression parameter 65 to process the compressed data to the inverse transform and the inverse transform, and generates a predicted differential signal value 67; The prediction difference signal value 67 is added to the in-plane prediction unit 69.卩73, the prediction image 71 of any of the pre-remaining parts 7G is output, the motion compensation 7昼4 image, 74; the in-plane prediction memory 77 is stored, the circuit filtering unit 78 is The decoded image 74 is subjected to filtering processing to generate a reproduced image 79; and the motion compensated prediction frame memory is used to store the reproduced image 79. The variable length decoding unit 61 is the moving image of the first embodiment. When the bit stream 60 is decoded and received, the bit stream 60 is subjected to entropy decoding, and the bit unit or the image block macro block size and the frame size formed by the image of one frame or more are decoded. In addition, in the case of the macro-region, the situation is not directly multiplexed into a bit stream and is defined by a profile, etc., based on the identification of the profile decoded by the bit stream in sequential units. E set block size. Based on the decoding block size of the decoding macroblocks of each frame, the optimal coding mode of each macroblock included in each macroblock is included in each frame. 62, material), pre-parameter parameter 63, compressed data 64 (that is, the quantized transform coefficient is equal to the compression parameter 65 (transform block size information, quantization step size) ,, the most decoded on the decoding side Good coding mode 62, best pre-323005 28 201143459 ^ number 63 I 枓 枓 64, the best compression parameter code device side coding best editing (four), best 18a, compression tribute 2 Bu optimal compression parameters. Exhaustion ==?7::r block size information is specified in the edit-change block size group ==:: r::change=: is the optimal coding mode - and the second s block size /..., #信号Specify the macro-block or sub-block transform inverse quantization-inverse transform unit 66, 压缩 you drag the compressed data 64 and the optimal pressure:: use the number: the specified block unit' Perform inverse quantization _ inverse transform processing, = = measure differential signal decoding value 67. Jt different output pre-addition, variable length decoding unit 曰, the system refers to the decoded peripheral When the movement of the block is decoded, the reference process determines the prediction vector, and by adding the difference value shown in FIG. 4, the motion vector is obtained: the decoded value = the predicted 1 movement of the stream 6° solution stone The vector raffinate includes a switch in which the money lion switching unit 68 switches the input destination of 63 in response to the optimum encoding mode 62. The switching unit 58 is input by the variable length decommissioning unit 61. The optimum prediction parameter 63 (in-plane prediction mode) is output to the in-plane prediction section 69, and when the optimal inter-frame prediction mode is displayed in the optimum 32350 29 201143459 coding mode 62, the optimal prediction parameter 63 (moving vector) is used. The identification number (reference image index) of the reference image indicated by each motion vector is output to the motion compensation prediction unit 70. The in-plane prediction unit 69 is stored by referring to the in-plane prediction memory 77. The decoded image (the decoded image signal in the frame) 74a in the frame generates a prediction image 71 corresponding to the intra prediction mode indicated by the optimal prediction parameter 63 and outputs it. The method of generating the prediction artifact 71 by the intra prediction unit 69 The operation of the in-plane prediction unit 8 on the side of the encoding device is the same, but the in-plane prediction unit 8 generates the prediction image 11 corresponding to all the in-plane prediction modes indicated by the encoding mode 7. The intra prediction unit 69 generates only the prediction image 71 corresponding to the in-plane prediction mode indicated by the optimal encoding mode 62. In this regard, the two are not the same. The motion compensation prediction unit 70 is based on the most input. The motion vector indicated by the good prediction parameter 63, the reference imaging index, and the like are generated by the reference image 76 of one frame or more stored in the motion compensation prediction frame memory 75, and are output. Further, the motion compensation prediction unit 70 generates a prediction image 72, and the motion compensation prediction unit 9 performs processing for searching for a motion vector from a plurality of reference images (corresponding to FIG. 3). In addition to the motion prediction side unit 42 and the operation of the interpolation artifact generation unit 43, only the processing for generating the prediction artifact 72 is performed in accordance with the optimal prediction parameter 63 supplied from the variable length decoding unit 61. Similarly to the encoding device, when the motion vector refers to the case of 30 323005 201143459 pixels outside the frame defined by the frame size, the motion compensation prediction unit 70 fills the pixels outside the frame with the pixels at the side of the frame. The method generates a predicted artifact 72. Further, the reference frame size is defined by the case where the size of the decoding frame is expanded to an integer multiple of the size of the decoding macro block, and the case where the size of the decoding frame is specified, and in the same order as the encoding device. To determine the size of the reference frame. The addition unit 73 adds any one of the prediction artifact 71 or the prediction artifact 72 to the prediction difference signal value 67 outputted by the inverse quantization-inverse transformation unit 66 to generate a decoded artifact 74. This decoded image 74 is used as a reference image (decoded image 74a) for generating an intra-frame prediction image of a subsequent macroblock, and is stored in the in-plane prediction memory 77, and is input to the loop. Filter unit 78. The loop filter unit 78 performs the same operation as the loop filter unit 27 on the encoder side to generate the reproduced image 79, and outputs it to the motion picture decoding device. Further, since the reproducing device 79 is used as the reference image 76 for the predicted image after generation, it is stored in the motion compensation prediction frame memory 75. In addition, the size of the reproduced artifact obtained after decoding all the macroblocks in the frame is an integer multiple of the size of the macroblock. In the case where the size of the reproduced image is larger than the size of the frame corresponding to the frame size of each frame of the image signal input to the encoding device, the reproduced image includes expansion in the horizontal direction or in the vertical direction. region. In this case, the decoded image obtained by removing the decoded image of the extended area portion from the reproduced image is output from the decoding device. In addition, in the case of the size of the reference frame, the size of the decoded frame is specified as 31 323005 201143459, and the 'decoding image of the reproduced image stored in the motion compensation prediction frame memory 75:: the decoded image' is predicted in the future. In the case of generation, *' may be formed as a solution to remove the extended area from the reconstructed image. The "I image is stored in the motion compensation prediction frame memory 75", which is known from X, according to the third embodiment. The structure of the moving image encoding device f is divided into macros corresponding to the encoding mode 7 of the macroblock/: the block 昼$5' will be preset to 3 according to the size of the huge money block or the sub-block. The plurality of transform blocks of the size of the block are changed, and the coding control unit 3 = transforms the block size and wealth, and the coding efficiency becomes the optimal one, and the block size is included in the optimal compression parameter 2Ga. The conversion-quantization is instructed: and the i-scaling unit 19 divides the optimal prediction difference signal 13a into blocks of the transform block size included in the optimal compression parameter 2〇a, and performs transformation and quantization processing to generate Compressed data 2 ^ Variable regions in large groups and informal ghost macroblocks or sub-block sizes herein are the conventional fixed method, that the code amount equivalent to improve the quality of encoded video. Further, the variable length coding unit 23 is configured by multiplying the size of the transform block that is adaptively selected by the deer coding mode 7 in the transform block size group, so that it corresponds to this. In the configuration of the moving image decoding apparatus of the first embodiment, the variable length decoding unit 61 decomposes the optimal compression parameter by the bit stream 6〇 in a macroblock or subblock unit. The deuteration-inverse transform unit 66 determines the size of the transform block according to the transform block size information included in the optimal compression parameter 65, and the block unit of the macro block size inversely transforms and dequantizes the compressed data 64. At 323 005 32 201143459, therefore, the 菱 区 大 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与 与Since the compressed data is decoded, the bit stream stream depleted by the moving picture decoding device can be correctly decoded. (Embodiment 2) (4) In the embodiment, the modification of the first embodiment, the modification of the worryable length coding unit 23, and the variable length of the dynamic image decoding device of the first embodiment are described. Modification of part 61

First, the second f_(four) variable length coding unit 23 will be described. The J f 9 diagram of the non-constructed code is shown in the internal H flat code device of the variable length chime 23 according to the first aspect of the present invention, and the first and second figures are omitted. The components are labeled with the corresponding component symbols, and the configuration of the device is the same as the component 2 of the second embodiment: the second variable length state is the same as the phase 2 variable length state of the embodiment. Illustrated in the above-mentioned first embodiment, the second embodiment is used. In addition, it is assumed to facilitate the mode group; = and: the code shown in Fig. 2A is used to use the == method. However, of course, it is also possible to adapt to the device composition of the processing method diagram read the frequency group: == long, part 23 contains: binarization table memory 表 table not correct mode 7 (or best prediction Parameter 10a 323005 33 201143459 and 18a, optimal compression parameter 20a) binarization table of the correspondence between the index value of the multi-value signal and the binary signal; the binarization unit 92 uses the binarization table to encode The optimum value of the multi-valued signal of the multi-valued signal selected by the control unit 3 (or the optimal prediction parameters l〇a and 18a, the optimal compression parameter 2〇a) is converted into a binary signal 103; arithmetic The encoding processing unit 1 converts the context identifying unit 102, the context information memory 96, the probability table memory 97, and the state transition table memory 98 generated by the reference context (CONTEH) generating unit 99, and converts the binarizing unit 92. The binary value 峨1〇^ is arithmetically encoded and the output coding bit is output. m, and the coding bit sequence m is multiplexed into the bit stream 30; the frequency information generation unit 93 calculates the optimal coding mode 7a (or the optimal prediction parameters i〇a and 18a, the optimal compression parameter penalty) The occurrence frequency* generation frequency information 94; and the binarization table update unit 95 updates the correspondence between the multi-valued signal and the binary signal of the binarization table of the binarization table memory (10) based on the frequency information 94. The variable length encoding order of the variable length encoding unit 23 will be described by taking the code control unit 3 code mode as an example. The optimum prediction parameter 10a for the parameter which is also the encoding target is explained. And l8a, the optimal compression parameter 2()a, the variable length can be edited in the same order as the best coding mode 7a. The j team, the m, and the m code control unit are set to the context. The (10) initialization flag 9 _ signal lQ 1 〇 二 二 二 二 二 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 323005 34 201143459 At the beginning of the dynasty, the flag was 91, and the context information memory initialization process was performed. S: The initial state is initialized by the initialization unit 90: the value ΠΜ 2 ' is the type of the best type of editing 7a that is stored in the reference binarization memory (10). The index value of the value of ί is converted into a binary signal 1〇3, and the arithmetic coding processing unit 1〇4 is rotated. The image of the output is saved in the memory table 105.

An example of the "encoding mode" shown in Figure 1 is the encoding mode (mh H nsm " shown in the table of "2h skip: the 2A skip: the total will be 3) added to the skip mode (曰 曰 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . In addition, the index values of these coding modes are binarized by up to 3 bits το and stored as "binary signals". In this case, the elements of the binary signal are referred to as rbin files. Further, in the example of Fig. 10, a small index value is assigned to a coding mode having a high frequency of occurrence, and the binary signal is also set to a short one bit, and the details will be described later. The optimum encoding mode 7a output from the encoding control unit 3 is input to the binarizing unit 92, and is also input to the frequency information generating unit 93. The frequency information generating unit 93 calculates this optimal encoding mode. The frequency of occurrence of the index value of the coding mode included in a (the selection frequency of the code mode selected by the coding control unit) is used as the frequency information 94, and is output to the 323005 35 201143459 binarization table update unit 95 which will be described later. The probability table memory 97 is a symbol of each bin file included in the storage table ^ γ 〇 ^ η , α tiger 1 〇 3 in order to keep the 碰 二 二 ^ 储存 储存 储存

Any symbol (MPS: M〇St Probable SymbGl) with a high probability of Probable S"h it is combined with its probability of occurrence. The probability value table of the probability table stored in the probability table memory 97 ("ll ^ ^, the system is 〇. 5 to 1-0 between the discreteness ut") is divided into a "probability table number".广矣Φ: The six-remembered body 98 is a memory that holds a storage table. The storage: reef, two * array is stored in the probability table memory 97. "Probability table II." "4" The code of the MPS in lj; the combination of the probabilistic state and the morphing of the money into the dragon state _. Figure 12 is a diagram showing the state transition of the state transition memory (10). The "probability table number", "[probability transfer of Μ code^", and "probability transfer after MPS coding" of the figure correspond to the probability table numbers shown in Fig. 11, respectively. 'When the probability of the probability table number ^ in the box, (from the 11th figure, the probability of occurrence of MPS is 〇.527), by using any symbol with a low probability of 0" or 1" (LPS: Least Probabie SymbQl) Riding code, probability (4) Migrating the shirt from "(10) coded probability migration" to probability table number Q (as seen from figure u, the probability of occurrence of MPS is 〇. 5〇〇). That is, by causing Lps to occur, the probability of occurrence of Mps becomes small. Conversely, when the MPS is encoded, the probability state indicates that the migration from "probability migration after Mps code 323005 36 201143459 code" is moved to the probability table number 2 (as shown in Fig. 11, the probability of occurrence of MPS is 0.550). That is, by causing MPS to occur, the probability of occurrence of MPS becomes large. The context generating unit 99 refers to the category signal 100 and the peripheral block indicating the types of the encoding target parameters (the optimal encoding mode 7a, the optimal prediction parameters 10a and 18a, and the optimal compression parameter 20a) input by the encoding control unit 3. The signal 101 generates the context identification information 102 for each bin file of the binary signal 103 obtained by binarizing the parameters of the encoding object. In the present description, the category signal 100 is the best coding mode 7a of the coding target macroblock. In addition, the peripheral block information 101 is an optimum encoding mode 7a of the macroblock adjacent to the encoding target macroblock. Hereinafter, the procedure for generating the context identification information by the context generating unit 99 will be described. Fig. 13(a) is a diagram showing the binarization table shown in Fig. 10 in a binary tree representation. Here, the coding target macroblock shown in the thick frame shown in Fig. 13(b) and the peripheral blocks A and B adjacent to the coding target macroblock will be described as an example.

In Fig. 13(a), a black dot is called a node, and a line connecting black circles is called a path. An indicator of the multi-valued signal of the binarized object is assigned to the terminal node of the binary tree. In addition, the depth of the binary tree corresponds to the bin file number, and the bit column of the symbol (0 or 1) assigned by each path from the root node to the terminal node is assigned to each other. The indicator of the multi-value signal of the terminal node corresponds to the binary signal 103. For each upper node (non-terminal node) of the binary tree, one or more context identification information is prepared in response to the information of the surrounding block A, B 37 323005 201143459. The C1, C2, 99, and the three context identification information of the selection CO, Cl, and C2 output the selected context identification information as the upper; = for example, in the figure 13(a), in the root node When three context identification information of c〇 is provided, the context generation unit adjacent to the peripheral block information of the peripheral blocks A and B and the identification information 102. 0 (The compilation mode of macroblock X is not 〇) 1 (The encoding mode of macroblock X is 0)

V C0 C1 C2 Γ (Α)+ Γ (B) = 0 Γ (Α)+ Γ (Β) = 1 Γ (Α)+ Γ (Β) = 2 (4) The above equation (4) is assumed as follows In the case of the base, the preparer, that is, the assumption that the peripheral block Α, Β is regarded as the macro block X, if the bat code pattern of the peripheral block 1 becomes "〇" (mb_skip), the coding target is huge. The probability that the block block will also become (mb_skip) will be high. Therefore, the context identification information 1() 2 selected by the above equation (4) is also based on the same assumption. Further, upper context nodes other than the root node are assigned i context identification information (Cl, C2, C3). The context information identified by the context identification information 102 is a value (0 or 1) and a probability table number approximating its probability of occurrence, which is now in an initial state. This contextual information is stored by contextual information memory 96 323005 201143459

The computing unit 104 refers to the context information 106 of the probability table number 97. Then, 06. Next, the arithmetic coding process determines the binary signal that is held in the context information arithmetic coding processing operation 1 to 3bin: 1〇6, and the MPS occurrence probability of the bin file corresponding to the probability table number 107 is 1〇8. Next, the arithmetic coding processing operation unit 104 performs the value of the MPS (G or 1) held by the context information 1〇6 and the determined (four) occurrence probability. The symbol value of the 档n file 0 is 1〇9 (0 or 1). Arithmetic coding. Next, the arithmetic coding processing operation unit refers to the state transition table memory 98, and obtains Mn according to the probability table number 1G7 held in the context information 1G6 and the symbol value 1〇9 of the bin file 0 which has been previously arithmetically coded. The probability table number 110 after the symbolization of the symbol. Next, the arithmetic coding processing unit 104 updates the value of the probability table number (that is, the 'probability table number 107) of the context information 1 〇 6 of the bin file 0 stored in the context information memory 96 to the probability table number after the state transition. (That is, the probability table number 110 after the symbol encoding of the bin file 0 previously obtained by the state transition table memory 98). Similarly to the bin file 0, the arithmetic coding processing unit 104 also performs arithmetic calculation based on the upper and lower 39 323005 201143459 f ^ information recognized by each context identification information 102, after the symbol encoding of each bin weight. , update the context information 106. The μ arithmetic coding processing unit 1〇4 outputs the coded bit sequence obtained by all the symbols of the Μn target ==, and the variable length coding unit 23 multiplexes the bit stream 30. On the iTfi: the context identifier identified by the context identification information 102 is L a n., and the parent 2 is updated by performing arithmetic coding on the *. That is, 'two. After the absence, the probabilistic state of the 卩 卩 & 会 会 : : : : 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 欠 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化 初始化9. The initialization unit 90 is the context information (slize) of the finger (4) part 3 of the Kang initialization flag 91 (4) = (4) bribery, the original pure is in the segment (a lifetime! 初始 initial of each context information 106 The state can be prepared in advance with a complex array, and the initial value of the number is edited. The state is also included in the context-based flagization binarization table update unit 95 value table update flag 113, and the data is controlled by the handle control unit. 3 indicates that the number of inflammations is equal to or lower than the index value of the optimal coding mode 7a in the household frequency # generated by the frequency information generation unit 93, and the binary value is updated by t 4 '. The table memory 105. The stomach is updated by the digitization table update unit 95 in this example, in accordance with the coding mode specified by the optimal coding mode 323005 40 201143459 7a of the coding object parameter. Frequency of occurrence 'to be able to occur with a shorter codeword (codeword) The highest coding mode is binarized' to update the correspondence between the coding mode and the index of the binarization table to achieve the purpose of reducing the code amount. Fig. 14 is an example showing the updated binarization table. The figure is an update state in a case where the state of the binarization table before the update is the state shown in the first diagram. The binarization table update unit 95 is in accordance with the frequency information 94, for example, mb_mode3. In the case where the probability of occurrence is the highest, the smallest index value is assigned in such a manner that the binary signal of the shorter coding character can be allocated to mb_mode 3. In addition, the 'binarization table update unit 95' has been updated in the binarization table. In the case of the case, it is necessary to generate the binarization table update identification information 112 required for recognizing the updated binarization table on the decoding device side, and multiplex it into the bit stream 30. For example, each encoding object parameter has a plural number. In the case of the binarization table, the IDs of the respective encoding target parameters are respectively assigned to the encoding device side and the decoding device side, and the binarization table updating unit 95 may be the ID of the updated binarization table. As a binarization table update The identification information 112 is outputted and converted into a bit stream. The control of the update timing is such that the encoding control unit 3 refers to the frequency information 94 of the encoding target parameter at the head of the segment, and determines that the frequency distribution of the encoding target parameter is greatly changed. When the predetermined tolerance range or more is exceeded, the binarization table update flag 113 is output for control. The variable length coding unit 23 multiplexes the binarization table update flag 113 into the segment header of the bit stream 30. In addition, the variable length coding unit 23 displays the coding mode, the compression parameter, and the prediction parameter when the binarization table update flag ^13 323005 41 201143459 displays "the update table is updated". In the binarization table, which binarization table has been updated, the binarization table update identification information 112 is multiplexed into a bit stream 3〇. Further, the encoding control unit 3 may update the binarization table at a timing other than the head of the slice, or may update the flag in the front wheel = the digitization table update flag 113 of an arbitrary macro block, for example. In this case, it is necessary to cause the binarization table updating section 95 to output the information of the macroblock position at which the binarization table update has been performed, and the variable length encoding section 23 causes the information to be multiplexed as the bit stream. 30. In addition, the encoding control unit 3 extracts the binarization table update flag to the binarization table update unit 95 to update the binarization table, and the context information initialization flag 91 It is output to the initialization unit 9A, and the initialization of the information memory 96 is as follows: 'The variable of the upper side is set to be the next, and the moving image decoding length decoding unit 61 of the second embodiment will be described. Fig. 15 is a view A block diagram of the internal structure of the variable length decomposing unit according to the second embodiment of the present invention. The configuration of the moving picture decoding apparatus according to the second embodiment is the same as that of the i-th: the first aspect, except for 0 Γ 4 E The shape operation is also described in addition to the t-th decoding unit, and the other components are the same as the catalogue, and the data is shown in Fig. 8. The variable length depletion unit 61 includes the rational operation unit m, which is a reference. Context generating unit 122 generates == ==, context information memory 128, probability table Memory = Migration Table Memory 135, which will represent 323005 42 201143459 'Optimal Encoding Mode 62 (or best prediction number 100, external ^b3, optimal compression parameter 65) that has been multiplexed into bit stream 60 The coded bit column 133 performs read decoding to generate a binary signal 137. • The binarization table memory U3 is assigned a binary signal table; the best: code mode 62 (or optimal prediction parameter 63, most The compression parameter 65) the binarization table 139 corresponding to the correspondence of the multi-value signals is stored; and the inverse binarization unit 138 uses the binarization table 139 to process the arithmetic solution 12 to generate the second The value 汛 137 is converted into a decoded value of the multi-valued signal 14 〇. Hereinafter, the entropy-decoded parameter is applied, and the optimal coding mode 62 of the macroblock included in the bit stream 6 为 is taken as an example. The variable length decoding order of the variable length decoding unit 61. The optimum prediction parameter 63 and the optimum compression parameter 65 which are parameters belonging to the decoding target are subjected to variable length decoding in the same order as the optimal encoding mode 62. The description may be omitted. Further, the bit stream of the second embodiment 60 includes: context initialization information 121, coded bit column 133, binarization table update flag 142, and binarization table update identification information 144 that are multiplexed via the encoding device side. The initialization unit 初始化2〇 initializes the context information stored in the above-mentioned “fight” 128, or the I is formed as: the initial state for the context information ( The initial value of the MPS value and the initial value of the probability table number approximating the + special rate is set in advance by the initialization unit 120 and selects an initial state corresponding to the decoded value of the context initialization information 121 from the group command. The context generating unit 122 refers to the parameter for displaying the decoding target (best 43 323005 201143459 'encoding mode 62, optimal pre-parameter information 126. Block Béch 24, and generates context recognition)

The category signal 123 is a parameter of the table reversal image, and the grammar pair stored in the variable length decoding unit 61 is H. Therefore, the butterfly device side and the decoding device side are stored here, and the code control unit 3 on the side of the editing device is kept. On the side of the encoding device, the variable length coding unit 23 is sequentially input (four) in accordance with the value of the type of the next coding parameter and its parameter (the finger value ^, that is, the class signal_). , the surrounding block | News 124 is the encoding mode obtained by decoding the macro block or sub-block, etc., in order to be used to decode the future macro block or sub-block... The story is first stored in the variable length solution=edge block information 124, and the memory of the upper and lower long text=6 is rotated as needed (not shown, the context generation unit 122 generates 120. Similarly, the context generating unit 122 on the decoding device side also generates the context identification information 126 in the binarization table 139 @per-bin file which is referred to by the inverse binarization unit 138. ^ Context φ in each bin file ^ The probability that the probability of occurrence of the MPS value (〇 or 1) == is maintained in the poor, and the probability information is used as the probability information for decoding the file. In addition, the probability table memory 131 and the state Migration Table Memory Coffee, 323005 44 ^1143459 • Department Storage and Editing The probability side of the device side is the same probability table (the 11th w 97 and the state transition table arithmetic decoding processing unit 127_) and the state transition table (Fig. 12). The coded bit column 133 is based on each bin w system. Converting the multiplex into the bit stream number 137, and outputting the inverse binary value to the arithmetic depletion to generate the binary arithmetic decoding processing operation unit i 8 body 128, and obtaining the information according to the editing stone position, 1 below Corresponding to each bin file of the 7° column 133 of the memory context identification information 126

The processing operation unit 127 refers to the probability table 139 °, and then the probability table held by the arithmetic calculus information 129: the memory 131, and the probability of occurrence 132. The MPS of each bin file corresponding to the humiliation 130 is followed, and the arithmetic decoding processing calculation unit 127 inputs the value to the arithmetic decoding processing operation based on the value of the MPS held in the context (0 or D, and the specified MPS occurrence probability 132). The coded bit column 133 of the unit 127 calculates and decodes the symbol value 134 (0 or 1) of each bin. After the symbol value of each bin file is decoded, the arithmetic decoding processing unit 127 refers to the state transition table memory 135. In the same order as the arithmetic coding processing unit on the encoding device side, the symbol of each bin file is decoded based on the symbol value 丨34 of each decoded wn file and the probability table number 130' held in the context information 129 ( The probability table number 136 after the state transition. The arithmetic decoding processing unit 127 then sets the value of the probability table number (that is, the probability table number 130) of the context information 129 of each bin file stored in the context information memory 128. The probability table number after the state transition is updated (that is, the probability table number 136 after the symbol decoding of each bin file previously acquired by the state transition table memory 135). The processing unit 127 rotates the binary signal 137 obtained by combining the result of the above-described arithmetic decoding with the sign of each bin file to the binary binarization unit 138. The inverse anti-binarization unit 138 is pressed by two. In the binarization table prepared for each type of the decoding target parameter stored in the value table memory 143, the same binarization table 139 at the time of encoding is selected and referred to, and is input by the arithmetic decoding processing unit 127. The binary signal 137 outputs the decoding target parameter code value 140. Further, when the type of the decoding target parameter is the encoding mode of the macroblock (optimal encoding mode 62), the binarization table 139 is the first map. The binarization table on the side of the encoding device shown is the same. The binarization table updating unit 141 updates the identification information 144 based on the binarization table update flag 142 and the binarization table update information 144 decoded by the bit stream 60. The update of the binarization table stored in the value table memory 143. The binarization table update flag 142 is information corresponding to the binarization table update flag 113 on the side of the encoding device, and is included in the bit stream 6标's header information, etc., to indicate whether the binarization table is updated or not. When the decoded value of the binarization table update flag 142 indicates that "the binarization table is updated", the binarization table update identification information 144 is further decoded by the bit stream 60. Binaryization table The update identification information 144 is information corresponding to the binarization table update identification information 112 on the side of the encoding device, and is information for updating the flag 113 for identifying the updated parameter on the encoding side. For example, As described above, when each of the encoding target parameters has a plurality of binarized 323005 46 201143459 tables in advance, the ID of the identifiable encoding target parameter and the id of the binarization table are respectively assigned to the encoding device side and the decoding device, respectively. On the side, the binarization table update unit 141 updates the binarization table corresponding to the ID value in the binarization table update identification information 144 decoded by the bit stream 6〇. In this example, the binarization table memory 143 is provided with two types of binarization tables and their IDs in the first and fourth figures, and assumes that the state of the binarization table before the update is the first map. In the case of the non-state, the binarization table update unit 141 performs an update in accordance with the binarization table update flag 142 and the binarization table update identification information 144.

In the case of 'the same thing', the binarization table corresponding to the ID included in the binarization table update identification information 144 is selected, so that the state of the updated binarization table is in the state shown in FIG. 14 and becomes the coding device side. The updated binarization table is the same. As described above, according to the configuration of the motion picture coding apparatus according to the second embodiment, the coding control unit 3 selects the optimum coding mode 7a, the optimum prediction parameters i〇a and 18a, which are optimal for the coding efficiency, and the best. The encoding target parameter of the parameter 2〇a is compressed and output, and the binarizing unit 92 of the variable length encoding unit 23 uses the binarization table of the binarization table memory 105 to represent the encoding target parameter represented by the multivalued signal. The frequency conversion processing unit 1〇4 performs arithmetic coding on the binary signal 1〇3 to output the coded bit sequence 111, and the frequency information generation unit 93 generates the frequency information 94 of the encoding target parameter. The binarization table update unit 95 updates the correspondence between the multi-valued signal of the binarization table and the binary signal based on the frequency information 94. Therefore, the conventional method of always fixing is compared to the binarization table. Reduce the amount of code based on the same quality of editing. ~ 323005 47 201143459 In addition, the binarization table update unit 95 causes the binarization table update identification information 112 indicating the presence or absence of the binarization table to be updated, and the binarization table update identification for identifying the updated binarization table. In the case where the information 112 is multiplexed into the bit stream 3', the moving picture decoding device of the second embodiment is configured by the arithmetic decoding processing unit of the variable length decoding unit 61. In the 127 system, the coded bit stream 133 of the bit stream is arithmetically decoded to generate a binary signal 137, and the inverse binarization unit 138 uses the binarization table 139 of the binarization table memory 143 to The value signal 137 is converted into a multi-valued signal to obtain a decoded value 丨4〇, and the binarization table updating unit ι41 updates the flag based on the binarization table decoded by the header information that is multiplexed into the bit stream 6〇. 142 and the binarization table update identification information 144 to update the predetermined binarization table in the binarization table memory 143. Therefore, since the moving picture image decoding device can update the binarization table in the same order as the moving picture coding device and de-binarize the coding target parameter, the motion picture coding device according to the second embodiment can be used. The encoded bit stream is decoded correctly. (Third Embodiment) In the third embodiment, in the moving image encoding device and the moving image decoding device according to the first and second embodiments, the prediction is performed by the motion compensation prediction of the motion compensation prediction neighbor 9. A modification of the generation processing of the artifact will be described. First, the movement supplement prediction unit 9 of the moving picture coding apparatus according to the third embodiment will be described. In addition, the configuration of the moving image editing device according to the third embodiment is the same as that of the first embodiment or the second embodiment, and the operation of each component other than the motion compensation prediction unit 9 is also 323005 48 201143459 In the same manner, Figures 1 to 15 are used in the following statements. The motion compensation prediction unit 9 of the third embodiment has the same configuration and operation except that the configuration and operation of the prediction image generation processing of the virtual sample accuracy are different from those of the first and second embodiments. In other words, in the second embodiment, as shown in Fig. 3, the internal illustrator generating unit 43 of the motion compensation predicting unit 9 generates reference image data of virtual pixel precision such as a half pixel or a 1/4 pixel. And practicing the virtual pixel precision

The reference image is used to generate the prediction image 45, by using the interpolation function of the six taps of six integer pixels in the vertical direction or in the horizontal direction as in the mpeg-4 AVC specification. The motion compensation prediction unit 9 of the third embodiment is enlarged by the super-resolution processing, and the reference image 15 of the integer pixel precision stored in the motion compensation prediction frame memory 14 is enlarged by the super-resolution processing. A prediction artifact is generated by generating a reference artifact 207' of virtual pixel precision and based on the reference artifact 2〇7 of the virtual pixel precision. - Next, the movement prediction unit 9 of the third embodiment will be described with reference to Fig. 3 . In the same manner as in the above-described first and second embodiments, the illustrator image generating unit 43 of the third embodiment is also moved by the motion compensation prediction frame memory Μ2 frame or more. 42 detects the movement within the predetermined movement search range on the reference 昼胄15 of =. The detection of the motion vector is performed by MpEG_4 in the same manner as the specification, and the motion vector of the pseudo pixel precision is checked. This detection method is a pixel information (called an integer pixel) held by the reference parameter, and a virtual sample (pixel) is created between integer pixels by interpolation operation 323005 49 201143459, and this sample is used as a reference image. The method to use. In order to generate a reference image of virtual pixel precision, a reference image of integer pixel precision must be enlarged (high definition) to generate a sample plan composed of virtual pixels. Therefore, in the case of the interpolation imaging reference image in which the virtual pixel accuracy is required, the interpolation imaging generating unit 43 of the third embodiment can use "w. τ Freeman, E. c. Paszt〇r and 0. T. Carmichael, "Learning Low-Level Vision,> , 1Ιη1:2ΓηΤΐ1〇ηΗΐ J〇Urnal 〇fComputer Vision, vol. 40, no. Test the super resolution technology to generate virtual pixel precision The reference image data stored in the reference, "Super Resolution", the reference image is entered into the image 2G7, and the motion detecting unit 42 is described using the configuration of the motion vector search process. The motion compensation pre-9 is the third embodiment of the present invention. Like the editing device diagram. Figure 16 (4) inside == raw f part = internal composition block = 2 ° 5, the system will move compensation prediction map = two = 15 to enlarge the processing; book him, with reference in 14 Like the reduction processing; the high _=^ part (10), the reference to the image 15 part 200 extracted high frequency 1 ^ 1, is the small end of the image processing fiber, the reference book like the ^ knife in the feature quantity; high frequency The feature output unit computing unit 202 calculates the feature quantity of the frequency range component; The correlation value between the quantities; the high-frequency component estimation unit of the same-frequency component pattern memory 204, 321005 50 201143459 to estimate the high-frequency component; and the addition unit to use the estimated high-frequency component correction to magnify the image The high score is generated, and the virtual pixel is generated as 偾9Π7. In Fig. 16, the reference image 15 for the range of the mobile search processing is input from the reference = image data stored in the motion compensation prediction map pivot memory 14. The interpolation image generation unit 43 is omitted, and the reference image 15 is divided into ^

The image reduction processing unit 200 and the high-frequency feature extraction unit 2〇1 J ftir The long image enlargement image reduction processing unit 200 is generated by the reference image 丨5, ^ 〇λα 赤'王 & 1/Ν (Ν is 2, 4, etc., the power of 2) is reduced in size, and the wheel feature extraction unit 201a. This reduction processing is performed by the general-purpose inter-frequency AA image reduction filter high-frequency feature extracting unit 201a, and the reduced image is formed by the image reduction processing unit 2, and the ancient step, the edge component, and the like are extracted. Feature amount. As the U feature amount, for example, parameters such as DCT or Wavelet transform coefficient distribution in the table can be used. The high-frequency feature extracting unit 201b extracts the high-frequency features similar to the high-frequency characteristics, and extracts the second feature amount having a different feature value from the reference image 丨5. The 2 feature quantity system ^::= The calculation unit 202 is also output to the high-frequency component estimating unit 2〇3. The correlation calculation unit 202 inputs the second feature day 3 from the high-frequency feature extraction unit 2〇ib into the local portion between the first calculation reference image 15 and the reduced image image by the high-frequency feature extraction unit 2〇h ★ feature amount. The block single ^ temple is the correlation value of the high frequency component range under the ten levy. In this case, the correlation value = special 323005 51 201143459 ‘The distance between the first feature quantity and the second feature quantity is, for example. The high-frequency component estimating unit 2〇3 specifies the high-frequency component from the high-frequency component pattern memory 204 based on the second feature amount input by the high-frequency feature extracting unit and the correlation value input by the correlation calculating unit 2〇2. The pre-existing equation, the reference artifact 207 that estimates and generates the virtual pixel precision should have: Frequency component. The generated high frequency component is output to the addition unit. ^ The image enlargement processing unit 2G5' is for the reference image 15 of the input person, and the blood is generated by the half pixel precision sample generation processing according to the MPEG-4 AVC standard only by being applied in the vertical direction or in the horizontal direction. An enlarged image in which the reference image 15 is enlarged to N times the vertical and horizontal directions by using an interpolation filter of six taps, an interpolation operation of a bilinear m, or the like, and an addition unit 206 is generated. The high-frequency component input by the high-frequency component estimating unit 2G3, that is, the high-frequency component of the magnified image is corrected, and the magnified image is magnified to N times the vertical and horizontal directions by the high-frequency component input from the high-frequency component estimating unit 2G3. Zoom in on the reference image. The interpolation key generation unit 43 uses the same reference image as the reference image of the virtual pixel precision with 1/N set to 1. Further, the internal illustrator image generating unit 43 may be configured to set an average value of 1/2 pixels or integer pixels by using N=2 to generate a half-pixel (1/2 pixel) precision reference image 2Q7. The interpolation operation of the filter is to generate a virtual sample (pixel) of 1/4 image accuracy. In addition, the interpolation imaging generating unit 43 may be configured to add the high-frequency component estimating unit 203 to the large image that is rotated by the imaging magnification processing unit 2〇5 in addition to the 16th configuration. The high-frequency component is carried out 323005 52 201143459 In the case of this configuration 2 = the result of the reference image 207 is generated. When the estimation accuracy of the effect of the adverse effect of the efficiency is deteriorated due to some reason such as the pattern specificity, the high-frequency component estimation unit suppresses the high-frequency component of the code that is selectively determined to be output. == High-frequency component = In-case case, etc. (4) Measurement, and the result of the two-image 45 is used to perform the motion-compensated display. Then, it is converted into a bit stream 30. The δ is mostly used as control information. Alternatively, the other parameters of the interpolation artifact generating unit (5) are determined unambiguously, and the control is performed as a bit stream. In the example determined by other parameters, the addition map of ~m6 or the compilation mode 7 shown in Fig. 2B. The class uses, for example, the motion compensation region block in the block (4) to report that when the macro is selected, the probability of belonging to the violent moving pattern is higher than that of the 码, = code mode, and the portion 43 is regarded as the super-resolution. The effect is very poor, and the interpolation image is added to the high-modulus portion 2〇6 outputted by the high-frequency component estimating unit 2〇3. On the other hand, when selecting a coding mode that indicates a macro:, a mode control, or a larger block size or a larger block heart: a compensation area type, it is a relatively static 昼 in-plane prediction. In the meantime, the internal illustrator image generating unit 43 considers that the image is controlled in a divided manner by adding a high frequency component estimation to the addition unit 较. The high frequency of the output of the jin is 323005 53 201143459 In addition to the use of the coding mode 7 u to take into account the size of the mobile vector, the parameters of the edge 乍 5, can also use the number. By adjusting the information of the deviation of the movement direction (4), the controller 30 ’ that does not directly process the addition can improve the compression efficiency. Yiyiwai* can be stored in the mobile compensation prediction frame memory 14, so that the mobile reward frame is more than 207, and then stored to form a virtual image precision reference image frame memory. The necessary ^ ^ ^ ^ on the mobile compensation pre-vector search and generate prediction book image;::: will add two, but moving:, reduce the motion compensation two not: == === degree of reference painting... generation : Silly 207* uses Figure 3 to show the reference gold = quantity = degree using virtual pixel precision; moving to _ * like ==? is used to generate and move the compensation area area image 41 in the preview mobile search range The inner image is compared to the predicted image 45 corresponding to the vector 44. The subtraction unit 12 generated by the prediction of the degree of motion 45 (predicted image 17) is transmitted by 4 degrees. The motion compensation area block image is called the salt method / 12 ' and is subtracted by » to become the prediction difference signal 13 . Code block image Signal 13 and integer pixel accuracy of the motion vector "° 3 (3) for the prediction difference 4 (predicted parameter 18) into 323005 54 201143459 line prediction efficiency evaluation. The evaluation of the prediction efficiency may be performed by the above formula (1) described in the above embodiment, and the description thereof is omitted. (Moving Vector Detection Sequence 11) The interpolation artifact generation unit 43' uses the motion vector 44 of the 精度 precision around the motion vector determined by the above-mentioned "moving vector detection I", as shown in Fig. 16 The interpolation artifact generation unit 43 is inside. The reference image of the virtual pixel precision generated by 卩 is generated to generate a prediction image 45. Similarly to the above-mentioned "moving vector detection 彳!", T' is a prediction image 45 generated by 1/2 pixel precision (predictive image (7), and the subtraction portion 12 is moved by the motion compensation region block image 41 (giant The set/subblock ^ is subtracted from 5) to obtain the predicted differential signal 13. Next, the encoding control section 1 points: No. 13 and 1/2 pixel precision of the motion vector "(predicted mobile yield estimation, and is located from In the motion vector of one or more 1/2 pixel precision around the two-direction I of the integer pixel precision, 2 the motion vector (moving vector detection order 111,) which makes the prediction cost L the smallest 1/2 pixel precision, The code control unit 3 and the motion compensation prediction unit 9 determine the motion vector of the 1/4 pixel precision, which is determined by the accuracy of 丄a彳冢京. The movement vector of the movement direction of 1/2 pixel precision: =, determines the prediction cost j, becomes the movement of the pixel precision to |44. ^ J ^ 174 (moving vector detection sequence IV,) Part 3 and the motion compensation pre-remaining unit 9 perform 323005 55 201143459 Detecting virtual pixels The motion vector of the degree is such that the motion compensation prediction unit 9 reaches a predetermined accuracy. The inner division becomes the coding mode 7 and the macro/sub-block image 5 is moved to the compensation region H block image 4 The meaning vector of the virtual pixel precision of the block unit of the unit (4) is determined by the predetermined precision, as the prediction parameter 18. The pre-'move generated by the prediction parameter 18 of the reference image indicated by the other quantity The compensation prediction unit 9 outputs the subtraction unit 12 to the subtraction unit: the image 45 (predicted image (7) is input to obtain the prediction difference signal 13. m macro/subblock image 5 subtraction 13 is output to the transform - The prediction difference outputted by the quantization unit 19' 12 is the same as the processing operation described in the above-described embodiment, and the third embodiment will be described in the same manner, and the description thereof will be omitted. The third embodiment is not limited. The image processing device according to the first and second embodiments, the configuration of the above-described second embodiment of the virtual pixel in the configuration of the domain of the upper-level prediction image generation process. The shape of the movement = the upper ^ other than due to the first figure It is the same as that of the decoding device. Therefore, in the second embodiment, the motion compensation prediction unit 70 is based on the reference image of the image of the + pixel or the 1/4 image. As in the MPEG_4AVC specification, a virtual pixel is created by interpolating two devices of six taps of six integer pixels in the vertical direction = ^ or in the horizontal direction to generate a prediction image, = The motion compensation prediction unit 7A of the third embodiment enlarges the reference image 76 of the pixel compensation precision by the motion compensation prediction frame memory 75 by the lexicographic process. Bucket > a like. Reference for generating virtual pixel accuracy 移动 The mobile fillet of the third embodiment? The 彳 竑 预测 预测 预测 预测 预测 预测 预测 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测 预测The motion compensation 9' test image identification number 7R ^ - the reference image memory 76 stored in the prediction frame memory 75 generates a prediction image 72 and outputs it. The accumulating unit 73 adds the moving supplement image 72 to the predicted difference weight value 67 in which the inverse quantization/inversion unit 66 input by the inverse quantization-reverse conversion unit 70 is inserted. A decoded image 74 is generated. In the operation of the motion compensation prediction unit 9 on the side of the encoding device, the motion detecting unit 42 that searches for the motion vector by the plurality of reference pictures is used. Inside: Book = Theory 4 is equivalent to the one shown in Figure 3 and is excluded according to the variable length solution 61 (4). Only the generated prediction image 72 = rationality (4) Jialai parameter 63, system ^, 'in When the virtual pixel precision is used to generate the predicted image ,, the reference number 70 of the reference gold silo=reference image specified by the n code (refer to the image search) is performed on the mobile patch prediction matrix 75. The image of Figure 16 is in the motion compensation prediction unit and the multi-docal image is used and the decoded motion, image 72 is used. In this case, when the code f is set to the side of the code: when the high-frequency component of the wide-frequency component estimation output is added to the amplification 1:, on the decoding h side, it is selected whether or not to extract from the bit stream (10). 57 323 〇〇 5 201143459 If the plug is set to have no addition processing, and the motion compensation time is determined by the parameters of other parameters unambiguously determined, the internal addition processing can be performed. In the same manner as the coding I of the moving direction, the variation of the profit field, and the like, and the movement of the surrounding area and the movement of the surrounding area are determined by the type _1 consumption prediction unit 70 The control information co-processed with the bat code device is more, and the code device side may not directly add the "movement compensation prediction to the reduction efficiency." Once the image is processed, it can also be only the reference 18a of the pseudo-pixel precision (ie, the best prediction parameter vector of the decoded side output indicates the virtual pixel precision). The prediction unit 9 also moves/compensates the prediction frame memory 14 in the case of the configuration: the generation unit 43 generates the virtual pixel precision 4:=: the intra-framed image reference image 15 or the virtual pixel image 207 to use. And the 17', the refined reference image 207 generates the predicted image body as: the 16th hidden large processing and the high frequency component are performed on the reference image stored in the shift_frame_body 75 Corrected virtual pixel master ^ ^ enemy time, should be prepared as a motion compensation prediction pattern. =: When the number of times is large, the pixel does not need to be sampled, so the amount of calculation can be reduced. Another ice & 16th In the case of the processing of the figure, the displacement range indicated by the other 'moving vector' has been pre-set 4nAA on the side of the solution 58 323 〇〇 5 201143459 code device, „ ^ 范围 粁筮 粁筮 的 也 也 也 移动 移动 移动 移动 移动It is limited to the manner of processing shown in Fig. 16 && θ displacement range, as long as the mold into the moving direction refers to the industrialization of the shift range of the shift finger, for example, the multi-device side and the de-flow 6〇 and transmitted, or in the application, in the Ma Shima If you want to do it, then, the decoding device will be set up. (4) The side of the decoding device is already based on the dynamic image encoding of the third embodiment, and the second two mobile compensation prediction unit... The reference image in the memory cell 并 and switch = component, and generate a reference pixel of virtual pixel precision: prime _ reference::=== virtual input image signal 丨 high frequency material 5 frequency component The reference of the frequency component: gentry, still can contain the majority of the image 17, and there is a 'ground compression two. The prediction of the motion compensation prediction generation. In addition, the motion picture of the implementation form of the Chu 〇 3 ^ mobile compensation prediction unit 7 〇 has丝佥^置' also generates virtual pixel production due to its composition order; the same smoothness of the moving 4 encoding device, whether it is multiplexed; bit stream; generating part, and cutting the reference book of predictive frame memory 75 Like the 76 second, use the reference image of the mobile complement to The pre- (4) virtual pixel is decoded by the method of the moving image of the third embodiment. Therefore, the bit stream of the encoding is correct. 323005 59 201143459 • In addition, the interpolation of the third embodiment described above The image generation unit 43 is configured to generate a virtual pixel precision image 207 by super-resolution processing based on the technique disclosed in WT Freeman et ai (2〇〇〇), however, The super-resolution processing itself is not limited to the second technique> test = other arbitrary super-resolution technology to generate a virtual image; the precision: the method of 207. Further, in the case where the coding apparatus of the 43rd embodiment is configured by a computer, the image coding control unit 3, the switching unit 6, and the motion compensation prediction frame memory 1 may be formed. =^^ The processing contents of the internal-inverse conversion unit 22, the variable-length coding unit, and the inverse-plane prediction memory 28 are as follows: the path filter unit 27, the memory of the computer, and (4) the c & image coding of the computer The program stores the image encoding program. Similarly to the storage of the memory, when the image decoding apparatus is configured by a computer, the dynamic code unit 6 of the third embodiment can be formed, and the inverse quantization-inverse conversion unit 66 can be changed. The length solution 69, the motion compensation prediction unit 70, the movement pre: 68, the intra prediction unit in-plane prediction memory 77, the circuit, the insect _ memory 75, and the image interpretation type are stored in the processing unit 78 of the computer. The moving image encoding program stored in the animation memory of the content is executed by the computer (10). [Industrial Applicability] The moving image encoding device of the present invention obtains an image-independent image, that is, the image is solved. Exhaustion device, because it can pre-set the macro block block large size 323005 60 201143459, it can still suppress the code amount of the total load such as the coding mode, and can efficiently compress the coded moving image coding device and the motion picture decoding. The device is suitable for a moving image encoding device that divides a moving image into a predetermined region and encodes the image in a region, and a moving image decoding device that decodes the encoded moving image in a predetermined area unit. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing the configuration of a moving image encoding apparatus according to a first embodiment of the present invention. Fig. 2A is a diagram showing another example of an encoding mode for performing an image of predictive coding in the time direction. Fig. 2B is a diagram showing another example of an encoding mode for performing an image of predictive coding in the time direction. Fig. 3 is a block diagram showing the internal configuration of a motion compensation prediction unit of the moving image coding apparatus according to the first embodiment. Fig. 4 is an explanatory diagram of a method of determining a predicted value of a motion vector in response to an encoding mode. Fig. 5 is a diagram showing an example of adaptation of the transform block size in response to the coding mode. Figure 6 is a diagram showing another example of adaptation of the transform block size in response to the coding mode. Fig. 7 is a block diagram showing the internal configuration of a conversion-quantization unit of the moving image coding apparatus according to the first embodiment. Figure 8 is a block diagram showing the configuration of a moving picture decoding device according to the first embodiment of the present invention. 61 323005 201143459 FIG. 9 is a block diagram showing the internal configuration of a variable length coding unit of the moving image coding apparatus according to the second embodiment of the present invention. Fig. 10 is a view showing an example of a binarization table showing the state before the update. Figure 11 is a diagram showing an example of a probability table. Fig. 12 is a view showing an example of a state transition table. Figure 13 is a diagram illustrating the generation order of context identification information; Figure 13(a) is a diagram showing a binarization table in a binary tree representation, and Figure 13(b) is a diagram showing a coding object macroblock and a peripheral area. A diagram of the positional relationship of the blocks. Fig. 14 is a view showing an example of a binarization table showing the updated state. Figure 15 is a block diagram showing the internal configuration of a variable length decoding unit of the moving image decoding device according to the second embodiment of the present invention. Figure 16 is a block diagram showing the internal configuration of a motion compensation prediction unit of the moving image coding apparatus according to the third embodiment of the present invention. [Description of main component symbols] 1 Input video signal 2 Block division unit 3 Encoding control unit 4 Macro block size 5 Macro/sub-block image 6 Switching unit 7 Encoding mode 7a Best encoding mode 8 In-plane prediction Part 9 Motion compensation prediction unit 10 Prediction parameter 10a Optimal prediction parameter 11 Prediction image 12 Subtraction unit 13 Prediction difference signal 13a Best prediction difference signal 62 323005 201143459 14 15 18 19 20a 22 24 26 28 30 40 41 43 45 50 52 54 61 63 65 67 69 71 Motion compensation prediction frame memory 73 Reference image prediction parameter conversion-quantization unit optimal compression parameter inverse quantization-inverse transformation unit local decoding prediction difference signal local decoding image signal in-plane prediction memory Volume bit stream motion compensation region division unit motion compensation region block key image interpolation image generation unit prediction artifact transformation block size division unit conversion unit quantization unit variable length decoding unit optimal prediction parameter optimal compression parameter prediction difference signal Decoding value, in-plane prediction unit prediction image addition unit 17 prediction image 18a optimal prediction parameter 20 compression parameter 21 compression capital 23 variable length coding unit 25 addition unit 27 circuit filter unit 29 local decoding image 42 motion detection unit 44 motion vector 51 conversion target block 53 transform coefficient 60 bit stream 62 optimal coding mode 64 compressed data 66 inverse quantization-reverse Conversion unit 68 switching unit 70 Motion compensation prediction unit 72 Prediction image 74, 74a decoding image 63 323005 201143459 75 Motion compensation prediction frame memory 76 Reference image 77 In-plane prediction memory 78 Circuit filter unit 79 Reproduction image 90 initialization unit 91 context information initialization flag 92 binarization unit 93 frequency information generation unit 94 frequency information 95 binarization table update unit 96 context information memory 97 probability table memory 98 state transition table memory 99 context generation unit 100 Category signal 101 Peripheral block information 102 Context identification information 103 Binary signal 104 Arithmetic coding processing operation unit 105 Binarization table memory 106 Context information 107 Probability table 5 Tiger code 108 MPS occurrence probability 109 Symbol value 110 Probability table number 111 Encoding Bit column 112 Binarization table update identification information 113 binary Table update flag 120 initialization unit 121 context initialization information 122 context generation unit 123 category signal 124 peripheral block information 126 context identification information 127 arithmetic decoding processing operation unit 128 context information memory 129 context information 130 probability table number 131 probability table memory 132 MPS occurrence probability 133 coded bit column 134 symbol value 135 state transition table memory 136 probability table number 64 323005 201143459 137 binary signal 138 inverse binarization unit 139 binarization table 140 decoded value 141 binarization table update unit 142 Binarization table update flag 143 Binarization table memory 144 Binary table update identification information 200 Image reduction processing unit 201a, 201b High-frequency feature extraction unit 202 Related measurement unit 203 High-frequency component estimation unit 204 South Frequency component pattern memory 205 Image enlargement processing unit 206 Adding unit 207 Reference pixel of virtual pixel precision 65 323005

Claims (1)

201143459 VII. Patent application scope: 1. A motion picture coding device, comprising: an coding control unit, which is a plurality of block division types to be a processing unit for motion compensation prediction or intra-frame prediction of an input image, The output specifies an encoding mode having one or more block division types. The block division unit divides the input image into a macroblock image of a plurality of blocks of a predetermined size, and is divided into the above-described coding mode. The block image of one or more blocks is output and outputted; the in-plane prediction unit is used to input the image of the block, and the image of the block is used for intra-frame prediction using the image signal in the frame. The motion compensation prediction unit generates a prediction artifact by using the reference artifact of one frame or more to generate a prediction artifact when the block imaging is input; And outputting the block artifact to one of the in-plane prediction unit or the motion compensation prediction unit according to an encoding mode of the block image output by the block dividing unit. The subtraction unit generates a prediction difference signal by subtracting the prediction artifact outputted by one of the intra-frame prediction unit or the motion compensation prediction unit by the block image outputted by the block division unit, and generates a prediction difference signal; And converting and quantizing the predicted difference signal to generate compressed data; and the variable length coding unit entropy encoding and multiplexing the coding mode and the compressed data to the bit stream; 1 323005 201143459 The foregoing coding control unit specifies that there is a predetermined:::= mode in which the selection root is a multi-value signal for output, and the bundle cutter domain-type compilation mode, the variable length coding unit has the following: ^Multiple values of mode: Part selected and pre-=, ^Edited:: Arithmetic coding processing operation part, :: Value: No. Apply arithmetic coding and turn out code bit 7 column to give multiple positions, and The coded binarization table is new: the bit stream; and the code mode selected by the coding control unit is each of the foregoing binarization table by the first 2 and the second value c. Edit d: portrait editing device, its + , update table of r-table: two values: two tables new more: department round-out table information multiplexed to bit element μ value table update flag header 3. If the patent application scope of the aforementioned variable length (4) The animation device described in the department, wherein the binarization unit uses a binarization specifying a correspondence between the compression b and the value §fl, and uses the multi-valued portion for conversion. And the various lesions are described above by the multi-valued signal of the transformation and quantization process. 323005 2 201143459 The reduced parameter is converted into a binary signal, and the binarization table update unit is transformed by the aforementioned compression parameter. - the frequency of use used by the quantization unit, updating the correspondence between the multi-value signal of the binarization table and the binary signal, and outputting a binarization table update flag indicating the update timing, wherein the arithmetic coding processing operation unit is The converted binary signal is subjected to nasal coding to output a coded bit column, and the encoded bit sequence is multiplexed with the aforementioned binarization table update flag to the bit stream. 4. The moving image encoding device according to claim 1, wherein in the variable length encoding unit, the binarization unit uses a correspondence between a multi-value signal indicating a prediction parameter and a binary signal. a binarization table of the relationship, wherein the intra-frame prediction unit or the motion compensation prediction unit is used for intra-frame prediction or motion compensation prediction, and the prediction parameter represented by the multi-value signal is converted into a binary signal, and the binarization is performed. The table update unit updates the correspondence between the multi-valued signal and the binary signal of the binarization table by using the frequency of use used by the in-plane prediction unit or the motion prediction unit, and outputting the representation. And updating the timing update table flag, wherein the arithmetic coding processing unit performs arithmetic coding on the transformed binary signal of the binarization unit to output a coded bit column, and the coded bit column and the second The value-added table update flag is multiplexed to the bit stream 0 3 323005 201143459. The movable image encoding device according to claim 3, wherein the variable length is The binarization table update unit of the degree coding unit outputs a binarization table update identification information for identifying the updated binarization table when the binarization table has a plurality of types, and the variable The length coding unit multiplexes the binarization table update identification information into a bit stream. 6. The apparatus according to claim 4, wherein the binarization table update unit of the variable length coding unit outputs a plurality of types when the binarization table has a plurality of types. The binarization table update identification information of the updated binarization table is identified, and the variable length coding unit multiplexes the binarization table update identification information into a bit stream. 7. A motion picture decoding apparatus, comprising: a variable length decoding unit that inputs, as a bit, a bit stream compressed and encoded in a macroblock unit of a plurality of blocks of a predetermined size; The elementary stream performs entropy decoding on the coding mode in the aforementioned macroblock unit, and entropy decodes the prediction parameter, the compression parameter, and the compressed data in a block unit that is divided according to the decoded coding mode; When the prediction parameter is input, the prediction target is generated by using the intra-plane prediction mode included in the prediction parameter and the decoded image signal in the frame; the motion compensation prediction unit inputs the prediction parameter. And using the motion vector included in the prediction parameter and the reference image specified by the reference imaging index included in the prediction parameter to perform motion compensation 4 323005 201143459 prediction, and generating a prediction artifact; The prediction parameters decoded by the variable length decoding unit are input to the aforementioned in-plane prediction unit in response to the encoding mode decoded as described above. Or one of the motion compensation prediction units; the inverse quantization-inverse transformation unit performs inverse quantization and inverse transform processing on the compressed data to generate a decoded prediction difference signal using the compression parameter; and an addition unit is used in the decoding prediction The differential signal is outputted with the prediction image outputted by one of the intra prediction unit or the motion compensation prediction unit, and the decoded image signal is output. The variable length decoding unit has an arithmetic decoding processing unit. The coded bit sequence of the encoding mode of the bit stream is arithmetically decoded to generate a binary signal; and the inverse binarization unit uses a correspondence between the binary signal and the multi-valued signal indicating the encoding mode The binarization table of the relationship converts the coding mode indicated by the binary signal generated by the arithmetic decoding processing calculation unit into a multi-value signal. 8. The dynamic image decoding apparatus according to claim 7, wherein the variable length decoding unit: the arithmetic decoding processing unit multiplexes the encoded bit sequence of the compression parameter of the bit stream Performing arithmetic decoding to generate a binary signal, and the inverse binarization unit uses a binarization table that specifies a correspondence relationship between the binary signal indicating the compression parameter and the multi-value signal, and performs the above-mentioned 5 323005 201143459 nasal decoding processing operation. The compression parameter generated by the part is converted into a multi-valued-value signal to represent it. 9. As described in the application for the hometown _7 item, the aforementioned variable length calculus horse system: In the arithmetic decoding processing unit, the coded bit sequence of the measured parameter is generated by arithmetic calculus, and the pre-binarization unit of the bit turbulence is used to specify the binary signal, the value signal and the multi-value signal. The second of the correspondence between the value signals and the arithmetic decoding unit calculated by the arithmetic decoding processing unit is converted into a multi-value signal by the aforementioned prediction parameters. To represent the dynamic image described in item 7 of the patent application scope: the variable length solution "has a binary table update flag on the information update of the information table on the update table of the binarization table. 11. The moving picture decoding device according to the eighth aspect of the invention, wherein the binarization table updating unit of the variable length decoding unit has a plurality of types in the binarization table, and is based on multiplex processing. The binarization table decoded by the header information of the meta stream updates the identification information to update the predetermined binarization table in the plurality of the binarization tables. The moving image decoding device according to the ninth aspect of the invention, wherein the binarization table updating unit of the variable length decoding unit has a plurality of types in the binarization table, The binarization table update identification information decoded by the header information of the bit stream is updated to update the predetermined binarization table in the plurality of the aforementioned binarization tables. The image decoding device according to claim 10, wherein the binarization table update unit of the variable length decoding unit has a plurality of types in the binarization table. And updating the identification information according to the binarization table decoded by the header information of the bit stream, and updating the predetermined binarization table in the plurality of the binarization tables. 7 323005