TW388843B - Moving image encoding method, moving image encoder and moving image decoder - Google Patents

Abstract

When an image is divided into a plurality of area for encoding, it has low degree of freedom in selection of area shape and difficulty in defining area shape according to characteristics of image. The moving image encoder has an area division section 1, a coding section 7, and a memory 9 that is used to predict motion compensation. The area division section 1 has a division processing section and an integral processing section. The division processing section divides an input image based on a reference relating to the propriety of division. The integral processing section integrates areas close to each other based on the reference relating to the propriety of integration. Then each area is coded. The integral processing diversifies the shape of the area.

Description

No. 86113874 Patent Specification Amendment Sheet Revised Date: 89 Greek 1.¾ fi ^ Record, only No. 86113874 Patent Specification Amendment Sheet Revised Date: 89 Greek 1.¾ fi ^ Catalog, Only A7 B7 Supplement V. Description of Invention ( 2) Output >> The adder 108 adds the decoded prediction error signal 107 and the prediction signal 102, and outputs it as a decoded video signal 109. The prediction unit 111 performs motion compensation prediction on the input image 1 and the decoded image signal 109 of the previous frame stored in the memory 110, and outputs a prediction signal 102 and a motion vector 112. At this time, this motion compensation is processed in a block unit of a fixed size of 16x16 pixels, which is called a macro block. For blocks in areas with more intense motion, it is also possible to use the method of dividing a giant block into 4 8x8 pixel sub-block units as an auxiliary function to process. The obtained motion vector 112 is sent from the $ image solution device, and the prediction signal 102 is sent to the difference 101 and adder 108. In this device, the motion compensation prediction process is used, so it is possible to compress the amount of data in the movie while maintaining the quality of the day. Fig. 28 is a block diagram of a video encoding device in the second conventional technique. This device is designed based on the coding method proposed by LCReal and others in "A Very Low Bit Rate Video Coder Based on Vector Quantization," (IEEE Trans, on Image Proceeding, VOL.5, No.2, Feb. 1996). In this figure, 113 is a region division unit, 114 is a prediction unit, 115 is a region determination unit, 116 is a coding mode data including inter-frame coding / intra-frame coding (inter / intra) data, and 117 is a motion vector. 118 is the encoding unit and 119 is the encoded data. In this device, the input image is first divided into a plurality of regions in the region division unit 113. The region division unit 113 determines the size of the region based on the motion-compensated prediction error. The region division Department 113 is pre-calibrated 5 * This paper size applies Chinese National Standard (CNS) A4 specifications (210 X 297 male «) I I-» ^ ills — ^ -1111111 1 ^. * (Please read the back first f write this page) Printed by A7 B7, Shellfish Consumer Cooperative of Intellectual Property Bureau, Ministry of Economic Affairs 5. Description of the invention (1) The present invention relates to a method and a device for encoding input animation, and a device for decoding the encoded animation. Round 27 is first In the conventional technology, a square circle indicates a structure of an animation encoding device according to ITU-T Recommendation H.263. In this figure, 1 indicates input digital shirt image data (hereinafter simply referred to as input image), and 101 indicates Differentiator, 102 indicates the prediction signal, 103 indicates the prediction error signal, 104 indicates the encoding section, 105 indicates the encoding data, 106 indicates the decoding section, 107 indicates the encoding prediction error signal, 108 indicates the adder, and 109 indicates the local decoded video signal. , 110 indicates the memory, 111 indicates the prediction unit, and 112 indicates the motion vector. 0 The input name image 1 to be encoded is first input to the difference 101. The difference J01 takes the difference between the input image 1 and the prediction signal 102. , — ......-----. Lose this as the pre-acceptance signal 103, and the editor's department l_〇f is the input image 1 for the name 41. Or The prediction error signal 103 is coded, and the coded data 105 is output. As for the coding method in the coding section 1 of the coding department, the roots,----i-Yin Jia, Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs (please read the precautions on the back before filling this page) According to the above proposal The method is to use an orthogonal transform DCT (Discrete Cosine Transformation) to transform the prediction error signal 103 from the spatial domain to the frequency domain, and linearly quantify the obtained frontal number of hits. Encoding data 105 It is divided into two directions. One is to send to the receiving end dragon image decoding device (not shown), and the other is input to the deblocking unit 106 in the device. The implementation of the decoding unit 106 is opposite to that of the encoding unit 105. , That is, the coded data 105 generates a fresh code prediction error signal 107. This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) No. 86113874 Patent Specification Amendment Page Revision Date: 89 Greek 1. ¾ fi ^ record, only amendment page of Patent Specification No. 86113874, date: 89 Greek 1. ¾ fi ^ record, stop A7 B7 Supplement V. Description of Invention (2) Output "Adder 108 is to decode the prediction error signal 107 and the prediction signal 102 are added and outputted as a decoded image signal 109. The prediction unit 111 performs motion compensation prediction on the input image 1 and the decoded image signal 109 of the previous frame stored in the memory 110. , The prediction signal 102 and the motion vector 112 are output. At this time, this motion compensation is processed in a block unit of a fixed size of 16x16 pixels, which is called a macro block. For blocks in areas with more intense motion, In other words, the method of dividing a giant block into four 8x8 pixel sub-block units can also be used as an auxiliary function for processing. The obtained motion vector 112 is sent by the $ image solution device. The prediction signal 102 is sent to the differentiator 101 and the adder 108. In this device, by using motion compensation prediction processing, the amount of animation data can be compressed while maintaining the quality of the day. Figure 28 It shows the structure of the image coding device in the second conventional technology. This device is based on LCReal et al. In "A Very Low Bit Rate Video Coder Based on Vector Quantization ,, (IEEE Trans, on Image Proceedi ng, VOL.5, No.2, Feb. 1996). In this figure, 113 is the region segmentation unit, 114 is the prediction unit, 115 is the region determination unit, and 116 is the interval between frames. Encoding mode data of inter / intra encoding (inter / intra) data, 117 is the motion vector, 118 is the encoding section, and 119 is the encoding data. In this device, the input image is first divided into the region division section 113 into Plural areas. The area dividing unit 113 determines the size of the area based on the motion compensation prediction error. The area division section 113 is pre-calibrated 5 * This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 male «) I I--» ^ ills — ^ -1111111 1 ^. *, (Please first (Read the back to write this page) Printed by the Shellfish Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, printed by the Shellfish Consumer Cooperative of the Central Sample Bureau of the Ministry of Economic Affairs, and printed by A7 B7. V. Description of the invention (3) 10 block sizes are prepared, Among 4x4, 4x8, 8x4, 8x8, 8x16, 16x8, 16x16, 16x32, 32x16, 32x32, the threshold value related to the degree of signal divergence between the frames is used for determination. Smaller blocks are used in areas with greater motion. In addition, in areas with small movements such as the background, larger blocks are used. According to Gu Chuan, the prediction error signal obtained by the prediction unit 114 is calculated by the divergence value in the area determination unit 115. Determine the block size. The area shape data, the attribute data 116 such as the coding mode, and the motion vector 117 are also determined at this time. Based on the coding mode data, the prediction error signal or the original signal is coded in the coding unit 118 to obtain coded data 119. The subsequent processing is the same as the first conventional technique. In the first known technique, the region shape of the coding unit is limited to 2 types and both are square V. Therefore, the coding suitable for the image frame structure or the characteristics of the image will be naturally limited. For example, if motion is required, When a larger subject improves its coding, although it is best to define a region of the same shape as the subject's range, this is difficult in the conventional technique. The second conventional technique is to provide regions of several sizes. From a block perspective, it is more critical than the first known technique. However, the device still confins the areas to a square. Therefore, even if there are already 10 kinds of squares, there is still room for improving the adaptability of image areas of arbitrary shapes. The present invention is based on the above-mentioned problems, and an object thereof is to provide animation coding technology for more flexible processing in accordance with the situation of an image to be processed. This paper size applies to Chinese National Standard (CNS) Α4 specification (210 × 297 cm) (Please read the precautions on the back before filling this page) ι ^ ϋ I i In · 1 1印 簟 A7 B7 V. Description of the invention (4) A more specific object of the present invention is to provide an animation coding technology, which adopts a region segmentation technology that can surely correspond to various image structures. Another object of the present invention is to provide a reference for segmentation according to various viewpoints when dividing a region to be encoded. Yet another object of the present invention is to provide a technology capable of accurately "encoding data for a region divided from various shapes of V, and performing correct decoding. The animation encoding method of the present invention includes segmenting according to a predetermined division judgment criterion. The steps of inputting the image as a plurality of regions; the steps of integrating the divided and plural regions according to a predetermined integration judgment criterion, and integrating the region and its surrounding regions; and the steps of encoding the image signals for the regions left after integration, respectively The above-mentioned judgment criterion for division or not is determined based on the comparison of the coding pros and cons of a certain region when the region is divided and when it is not divided. The above-mentioned criterion for integration judgment is based on a certain region, its region and its surroundings. The result of the comparison between the compilation and the current when the region is integrated and when it is not integrated is determined. 'On the other hand, the animation encoding device of the present invention includes an area division section and an encoding section. The area division section includes a division processing section for Divides the input image into The plural area and an integration processing unit are used to judge the plural areas divided by the division processing unit respectively according to a predetermined integration judgment criterion, and the integration area and the surrounding area are integrated. The coding unit controls the integration area by the integration processing unit. The areas left by the integration are coded separately for their image signals. The paper size applies to the Chinese National Standard (CNS) A4 specification (210X297 mm) τ --..------ installation ------ order-- ---- '/ Λ · * (Please read the notes on the back before filling this page)

Printed by the Ministry of Economic Affairs of the Bureau of Standardization, Shellfish Consumer Cooperative, I Α7 Β7 V. Description of the Invention (5) The above-mentioned integrated processing section includes a temporary coding section for preliminarily encoding images for each area and calculating its code, A decoding section for decoding an image encoded in the temporary encoding section, a coding distortion calculation section for calculating encoding distortion using the image decoded by the decoding section, and calculating both An evaluation value calculation unit that judges the evaluation value of the coding quality. For each area, based on the comparison of the evaluation value obtained in the case of integration with the surrounding area and the evaluation value obtained in the case of unintegration, 'decide whether to perform the integration of the area / The above-mentioned division processing section includes an action calculation section for calculating the judgment The prediction error power of the motion compensation prediction with each area is used as the motion parameter of the area, and a division judgment section is used to compare the calculated motion parameter with a preset reference value. When the comparison result is that the motion parameter exceeds the reference value In the case of this, the region is divided into a more detailed region β or the above-mentioned segmentation processing unit includes an action calculation unit for calculating the edge strength of the original signal of each region as an action parameter of the region, and a A division judgment unit is used to compare the calculated motion parameter with a preset reference value. When the comparison result is a case where the motion parameter exceeds the reference value, this area is divided into more detailed areas. Alternatively, the above-mentioned segmentation processing unit includes an action calculation unit for calculating a linear sum of a plurality of values representing the image characteristics of the region for each region, as the action parameters of the region, and a segmentation judgment unit for Compare the calculated motion parameters with preset reference values. When the comparison result is that the action parameter exceeds the reference value, this area is divided into more detailed areas. This paper size applies to China National Standard (CNS) Α4 (210 × 297 mm)-'* I i:-! H n I "' (Please read the notes on the back before filling this page) Set the central standard of the Ministry of Economic Affairs扃 Beigong consumer cooperation Du Yince A7 B7 V. Description of the invention (6) The above-mentioned plural values may also include the code amount and prediction error power of the motion parameters of each region in the motion compensation prediction. In addition, the above-mentioned plural values may also include the code amount of the motion parameters of each region, the prediction error power in motion compensation, the divergence value of the original signal, the edge strength, and the size of the motion parameters of each region. The above-mentioned segmentation processing unit further includes a hierarchy identification unit for determining the importance of each area as a hierarchy parameter, and judging whether each area can be divided from the action parameters and the importance. The above-mentioned hierarchy recognition unit observes the structure of the subject across a plurality of regions, and determines the hierarchy of each region. The above-mentioned subject structure can also be judged by the original signal divergence of the area, the edge attack, and the program connected to the surrounding area on the edge. In addition, the above-mentioned level recognition unit observes the feature amount of the image to detect the subject Based on the results, determine the level of each area. At this time, the above-mentioned hierarchical recognition department sets up a subject included in the image in advance, and stores in advance image feature values including the subject, and stores the feature values of each area image and the M The consistency of the characteristics of the subject determines the level of each area. The above-mentioned segmentation processing unit includes a temporary encoding unit to preliminarily encode an image for each region and calculate a code amount thereof, a decoding unit to decode the image encoded in the temporary encoding unit, and an encoding calculation using the image decoded by the decoding unit. Distortion coding distortion calculation unit and 9 paper sizes that are used to calculate the evaluation value for judging coding quality considering both code size and coding distortion. Applicable to China Paper Standard (CNS) A4 (210X297 mm) ( (Please read the precautions on the back before filling out this page)

、 1T M Printed by the Central Bureau of Standards and Quarantine of the People's Republic of China, Shellfish Consumer Cooperative A7 _____B7 V. Description of Invention (7) Evaluation value calculation department. For each area, a decision is made as to whether to divide the area based on the comparison between the evaluation value obtained when it is divided into smaller areas and the evaluation value obtained when it is not divided. In addition, "in the temporary editing unit", the quantization parameter of the prediction error signal accompanying the motion-compensated prediction is dynamically set, and the "evaluation value calculation unit calculates the evaluation value while changing the quantization parameter." Furthermore, an evaluation value calculation section is provided in the previous paragraph of the temporary encoding section, to obtain a linear sum of the code amount of the motion parameters of each region and the prediction error power in the motion compensation prediction as the evaluation value. The above-mentioned temporary encoding unit can also detect motion parameters based on the evaluation value. On the other hand, the animation decoding device of the present invention receives encoded data of an image encoded after being divided into a plurality of regions. It includes an area shape decoding unit for restoring the shape of each area divided during encoding based on the area shape information contained in the encoded data; and an image data decoding unit for confirming the area based on the restored area shape. The encoding sequence is used to decode the image of each area from the encoded data. At this time, the shape information of the area includes information related to the process of dividing and integrating each area during encoding. Based on this information, the shape decoding unit of the area uses the same processing as the encoding device to reproduce to confirm the division status of the area. Embodiment: [First Embodiment] The first circle represents the structured block circle of the animation encoding device in this embodiment. This device can be used in, for example, video calls or video conferences. It is used for 10 paper sizes that apply the Chinese National Standard (CNS) A4 specification (210X297 mm) (please read the precautions on the back before filling out this page). Standards of the Ministry of Economic Affairs, Ai Binggong Consumer Cooperatives Co., Ltd. A7 B7 V. Description of Invention (8) 'Portable or fixed equipment for image communication. In addition, it can also be used as a video encoding device for video casting and recording devices such as digital VTRs and video servers. Even the processing program in this device can be applied to an animation encoding program implemented in software or firmware of Digital Signal Processing (DSP) " In the first figure, 1 is Rotate the image, 2 is the area segmentation part, 3 is the area shape data, 4 is the area image data, 5 is the area motion data, 6 is the area attribute data, 7 is the encoding part, 8 is the locally decoded image, 9 is the memory, 10 is a reference image and u is a coded bit string. Fig. 2 shows the operation flowchart of the device. First, the overall operation of the device will be described with reference to FIGS. 1 and 2. The input image is input to the area division section 2 (S1), where it is divided into a plurality of areas. The area segmentation unit 2 is used to perform two processing steps such as the initial segmentation (S2) and the surrounding area integration (S3) described later. The area segmentation unit 2 applies shape data 3 and image data to each area obtained by the segmentation result. 4. Sport data 5, coding mode of each area, etc. 'attribute data 6 are transferred to the coding section 7. The encoding unit 7 outputs these data conversion processing and multiplexing into a bit pattern 'according to a predetermined encoding method as an encoded bit string n (S4, S5). Before the last area, the encoding of each area is continued (S6, S7). In addition, in order to perform the process of area division and encoding based on motion compensation prediction, the encoding unit 7 generates a locally decoded image 8 of each area and Stored in memory 艎 9. The area division unit 2 and the decoding unit 7 take out and store the paper in the chronological paper, and apply the Chinese National Standard (CNS) A4 specification (2 丨 0 × 297 mm) ΙΊ--Γ ---- i (l ^ .-- •-( (Please read the precautions on the back before filling this page),? Τ The local consumerism cooperation agreement of the Ministry of Economic Affairs of the Central Bureau of quasi-industry Du Yin 11 A7 B7, invention description (9) Local decoded image of memory 9 as reference image 10, Motion compensation prediction. Figure 3 shows the detailed structural circle of the area cut section 2. In this figure, 12 is the segmentation processing section # 13 is the initial segmentation shape data, and 14 is the integration processing section. (1) The initial segmentation corresponds to the first The initial segmentation step of S2 in 2 circles is performed in the segmentation processing section 12. The initial segmentation refers to the segmentation processing performed before the integration. The total number of segmentation is determined by the state of the image, that is, the characteristics or characteristics of the image The fourth circle shows the internal structure of the segmentation processing unit 12. In this figure, 15 is a uniform segmentation unit, 16 is an activity calculation unit, 17 is an action parameter, 18 is a segmentation judgment unit, and 19 is a segmentation status indicator. Signal. The so-called action is to judge the image The feature or characteristic is quantified data determined according to the established properties. The action here is to use the prediction error power accompanied by the motion compensation prediction of the area. Figure 19 shows the motion compensation by block matching. Prediction method. In the block matching method, the vector v obtained by the following formula is used as the motion vector of the region S to be predicted.

Dmin = min (Σ tf s (x + vx / y + v. T-i)-fs (x, y, t)]) veR g fs (x, y, t) is the predicted region S in time t The pixel value at (x, y), fs (x, y, tl) is the pixel value at (x, y) in time t_l, and fs (x + vx, y + Vy, t · 1) then Is the pixel value at the position (x, y, tl) shifted by the vector v "In addition, R represents the motion vector search range. This paper size applies Chinese National Standard (CNS) A4 specification (210X297 cm) -i ^ i · al ^ i ^^^ 1 n ftjl ——I— m I • I (please read the note ^^ on the back first) (Fill in this page) Order A7 printed by the Ministry of Economic Affairs Central Standards Bureau Bureau of Consumers and Cooperatives ------- a -______ B7_ V. Description of the invention (10) Based on the results obtained by the director, the predicted image is ^ x + Vx'y + v ^ -o 'prediction error power, that is, the action "is °. In this method," the method can be used to perform a region based on the local motion complexity of the image with the defined action parameters. Segmentation. Therefore, in the control such as knitting, it is possible to subdivide the part with more severe motion and coarsely divide the part with smaller motion. In addition, similarity motion compensation and check can be used to obtain similarity (affine) motion parameters. Perspective motion compensation for 3 degree spatial motion, etc. Fig. 5 shows the operation flowchart of the segmentation processing section 12. In this figure, the first uniform segmentation section 15 performs uniform block segmentation without specific conditions (S8). , For example, as shown in the sixth circle, a picture frame is divided into blocks of 32x32 pixels " > This segmentation process is called the 0th segmentation stage. The number of blocks generated in the 0th segmentation stage is represented by B0n (1 1n SN〇) for each block, and then 'individually judge whether each BGnS needs to be updated. Further block segmentation is performed (S9). To this end, the motion parameter 17 of each B0n is calculated in the motion calculation section 16. The segmentation determination section 18 performs a preset preset value T0 and the motion parameter of each block. In comparison, when the action parameter 17 is larger than T0 ',' This BGn is divided into 4 equal divisions (S10). This proceeds to the last block (S11, S12) 'This is the first division stage.

I Figure 7 shows the image segmentation state at the end of the first segmentation phase. The number of newly generated 16x16 pixel blocks is Nm, and each block is represented by BUlSMNi). After that, calculate the operating parameters of each B and 13 paper sizes that are applicable to China's national standard (CNS > A4 size (2 丨 0x297 mm) (Please read the note on the back before filling this page)

Printed by the Central Bureau of Standards of the Ministry of Economic Affairs-Industrial and Consumer Cooperatives A7 B7 V. Description of the invention (11) and use the threshold value TH1 for the second division stage. Then, for the j-th partition stage, the threshold THj is adopted, and the j + 1th partition stage is implemented (S13 ~ S16). When the value of j reaches a predetermined upper limit value, the initial division is ended. The explanation here is based on the second segmentation phase. At this time, the block shown in Fig. 8 is finally generated. The block size can be 8x8 pixels to 32x32 pixels. At the end of the initial division, the number of blocks is M0, and each block is represented by an initial block sn. The shape data of SON is transmitted to the integration processing unit 14 as the initial segmented shape data. (2) Integration of surrounding areas Next, the integration processing unit 14 integrates the surrounding areas of each s0n. The internal structure of the integration processing unit 14 is shown in Section 9_ In this circle, '20 is the labeling unit '21 is the surrounding area setting unit, 22 is the temporary editing unit, 23 is the decoding unit, and 24 is the editing unit Ma distortion calculation section, 25 is an evaluation value calculation section, 26 is a constant used to calculate the evaluation value, 27 is an integration judgment section, and 28 is an integration processing cycle instruction signal. FIG. 10 shows an operation flowchart of the integration processing unit 14. In this circle, 'in accordance with a certain rule, the marking unit 21 first assigns a number to the initial region 80 (1), which is a label (lebel). For example, the pixel unit in the image frame is performed from the upper left to the lower right corner. Scan, and sequentially assign each area number. Figure 11 shows a simple example of marking. In this figure, according to the order presented on the scanning line, each block is marked with r 1 "," 2 "... At this time 'No Consider the size of the area. Below, the label value of the area Skn is represented by l (Skn). In addition, this k value corresponds to the k-th integration stage described below, and k = 0 in the initial state. This paper scale applies to Chinese national standards (CNS > Α4 specification (2 丨 0X297) 嫠 JT .-------- cr--(Please read the precautions on the back before filling in this page) Industrial and consumer cooperatives printed A7 £ 7_ V. Description of the invention (12) Next, in the surrounding area setting section 21, the "surrounding area" (S18) of each area is defined by a mark. Circle 12 is an example of the surrounding area. According to The mark in the first figure indicates the circle area of the area So. That is, the areas B, C, and D are connected to the edges of the area A as the target, and the marker value is larger than the area A as the target, so the trace is the surrounding area. Then go to each area to determine whether This area can be integrated with its surrounding area. To this end, the temporary encoding section 22, decoder 23, encoding distortion calculation section 24, and evaluation value calculation section 25 are used to calculate an evaluation value for integration (S19). The evaluation value is as follows The relationship between the code shown in the formula and the distortion cost (c0st) "I (5 *) = D (5 *) + AR (5 *) (where D (Skn) is the coding distortion of Sfcn, which is the square The sum of the errors 'R (Skn) is the code amount of Skn, and λ is a constant 26. Integration is performed in the direction that L (Skn) becomes smaller ^ L (Skn) becomes smaller, which is equivalent to the given constant' Reduce the encoding distortion within the predetermined code size range. By reviewing the reduction of L (Skn>) in the picture frame, you can reduce the encoding distortion under the same conditions. Figure 13 is a detailed flowchart of S19. First, Prepare the Skn encoding in the temporary encoder 22 (S22). The purpose of this encoding is to prepare to calculate the code size R (Skn) and derive the encoding error. D (skn). In this embodiment, the temporary encoding unit 22 uses the reference image 10 to perform motion compensation prediction. The encoded data includes image data, that is, the prediction error signal or the original signal, and is used to specify the motion data of the predicted image. , And coding attributes, etc. (Read the precautions on the back before filling in this page)-Binding · Order 15 A7

V. Description of the invention (13) The policy of the Shellfish Consumer Cooperative of the Central Bureau of the Ministry of Light Economy, the sum of these code amounts is R (Skn). The prediction error signal is obtained from the difference between the original signal of the region Skn and the predicted image. On the other hand, the decoded data obtained by the temporary encoding unit 22 in the decoding unit 23 generates a locally decoded image of Skn (S23). Next, the editing distortion calculation unit 24 calculates the distortion D (Skn) of the local demapping image and the original image (S24). The evaluation value calculation unit calculates a code amount-distortion cost L (Skn) based on R (skn) and D (Skn) (S25). In step S19, three evaluation values are calculated for all regions according to the calculation method of the above evaluation values: 1. Each region Skn itself ... L (Skn) 2 · Skn's surrounding region Ni [Skn] ... L (Ni [Skn]) 3 · Suppose the area where Skn and Ni [Skn] are integrated ... L (Skn + Ni [Skn]) where Ni [Skn] represents the surrounding area of! ^ Η and i is the number used to distinguish the surrounding areas. The integration judging unit 27 searches for the position where the value of Dl = Ι (5 *) + Ι (^ [^])-Ι (5 * + ^ [5 *]) is the largest in the image frame, and this Skn and Ni [Skn] is integrated (S20). This is the k-th integration segment "After that, the integration judgment unit 27 instructs the marking unit 20 to update the mark through the integration processing cycle instruction signal 28. The tree section 20 replaces the mark 1 (Ni [Skn]) with the mark 1 (Skn), and sets the surrounding area with the peripheral area setting section 21. A new region Sk + 1n and a surrounding region NKSk + 1n) are obtained here, and L (Sk + 1n), L (Ni [Sk + 1n]), and L (Sk + 1n + Ni [Sk + 1n]) are obtained. The integration judging unit 27 stops the instruction to the marking unit 20 when no combination can produce a positive DL value, and ends the integration process. 16 This paper size applies the Chinese national standard (CNS > A4 specification (2 丨 0X297)} (Please Please read the notes on the back before filling in this page) n ^ n— ^^^ 1 1 ^ 1 —ι >. I. Ordered by the Ministry of Economic Affairs, Bureau of Labor, Cooper ’s Consumer Cooperative, A7 B7, V. Description of the invention (14 ) ~ '~ (S21) 〇 After the above processing of region segmentation and integration is completed, data indicating the region segmentation status of the last input image 1 3, image data of each region 4, motion data 5, attribute data 6 rounds It goes to the encoding section 7. After that, encoding is performed by a predetermined encoding method. 'In this embodiment,' not only segmentation, but also integration. Therefore, in the end, each area can be presented as a set of square blocks of various sizes. For example, for a subject with large motion in the image, a single area similar to its contour line shape can be integrated. As a result, the quantization parameter of each subject can be changed to control the code amount, which flexibly corresponds to the actual Shadow In addition, under the limitation of the provided code amount, the most appropriate region division with the smallest coding distortion can be presented. Therefore, compared with the conventional _animated inscription device, it is easy to implement with less code amount High image quality. In addition, 'Although the initial segmentation is completed at the second segmentation stage in this embodiment, it can also be completed at other stages. For example, when the movement of the image trimming is small, you can At the end of the first sub-cut phase, the number of phases can also be increased in the opposite case. In addition, although the image frame is used as the coding correspondence in this implementation, it can also be applied to any shape in the image frame. The subject is a rectangular image surrounded by an external square. In the encoding unit 7 and the temporary encoding unit 22 of this embodiment, although the area skn is encoded by a combination of DCT and linear quantization, other Methods, such as vector quantization, sub-band coding, this paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 1 ^ ---- 111 ^ ί 丫 装-• · ( please Read the notes on the back before filling this page.) Order the A7 B7 seal of the Bayong Consumer Cooperative of the Central Ministry of Economic Affairs of the Ministry of Economic Affairs. 5. Description of the invention (15) The wavelet encoding method can also be used. It can also be Provides several encoding methods and selectively uses the best encoding efficiency structure. Although the prediction error power is used as the action parameter in this embodiment, it can also be considered in other examples, for example: First An example is the dispersion value in a region. The dispersion value indicates the complexity of the pixel distribution in the region. In an image region that contains sharp changes in pixel values such as edges, the divergence value assumes that the pixel value of the region S is fs (x, y, t), the average pixel value in the region S is Ps, and the divergence value can be obtained by the following formula. σϊ = go ~) 2

If Ns uses this action parameter, it can divide the area according to the structural complexity of the image area. Therefore, it can control the encoding of the details of the parts with sharp changes in pixel values and the rough encoding of the parts with less changes in pixel values. The second example is the edge strength in the area. "Edge strength can be used as described by G. Robinson in" Edge detection by compass gradient masks "(Journal of Computer Graphices and Image Processing, Vol. 6, No. 5, Oct. 1977). The recorded Sobel operator is obtained as the number of pixels (edge distribution area) distributed on the edge green. In the case of this method, the region can be divided according to the edge green structure of the image, so it is possible to control the encoding of the arsenic part with the edge green and the rough encoding of the part without the edge green. The third example is the magnitude of motion parameters predicted based on the motion compensation of the region. The result of motion compensation prediction is to find motion parameters. In this paper, the paper size of this paper applies the Chinese National Standard (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before filling this page). Β7 V. Description of the invention (16) In the case of the matching method, 'is the vector v. In this method, the block can be divided according to the degree of image movement, so it can control the production and locality of the subject and so on. Partial detailed coding for larger motions, rough coding for parts that have almost no motion, such as background regions. The fourth example is the code amount of the operation parameters predicted based on the motion compensation of the region, and the linear sum of the power of the prediction error. The evaluation value of the situation is defined as follows: L «nc = Dmc + and Rmc. (Equation 2) \ Here Dmc is the prediction error power obtained during the detection of the motion parameter 'λ is constant and Rmc is motion The code size of the parameter. Lmc is reduced to obtain the motion parameter, and the evaluation value at this time is used as the motion parameter. In this method, the area is divided into the amount of data and the complexity of the motion complexity through the image. The total encoding cost of the parameter data amount is reduced, and the region encoding can be performed with less data amount. The fifth example is the sum of the action values described above. Borrow Since various motion values are appropriately weighted, various kinds of images can be processed. [Second Embodiment] This embodiment Λ relates to a partial deformation device of the region dividing unit 2 in the first embodiment. The fourteenth circle is this embodiment. The internal structure circle of the middle region division unit 2. In this figure, the region division unit 2 of this embodiment has a uniform division unit 15 instead of the division processing unit 12 of FIG. 3. In this structure, as shown in FIG. As shown in the figure, in the initial segmentation process, the critical value of the operating parameters is not judged, but the minimum area is unconditionally divided into regions. The paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) ^ ϋ · IV HI In n _ · (Please read the notes on the back before filling out this page) Order the A7 __B7____ printed by the Central Laboratories of the Ministry of Economic Affairs of the People's Republic of China. 5. The square block in the area of the description of the invention (17). This Smallest The area area can be set. In this example, it is not necessary to set a critical value, so only the code size-distortion value is used as the evaluation value to divide the area. Therefore, neither the steps related to setting the critical value nor the The calculation of the action parameters and the processing of comparison and judgment are not needed. Therefore, when the calculation burden on these processes needs to be reduced, this embodiment can be used instead of the first embodiment. [Third embodiment] In this embodiment, In the segmentation process, when determining whether it can be segmented, it includes not only action parameters, but also indicators indicating the importance of the area (called a level, class). On areas of high importance, it is best to use detailed coding. And the area is smaller. The less important areas are made as large as possible to reduce the code size of each pixel. For example, the action parameters are closed, local statistical directors within a region. On the other hand, the hierarchical parameter in this embodiment is based on the characteristics of the cross-region image. In this embodiment, the hierarchical parameter ′ is defined based on the structure of the object across the area, to what extent a person can see, that is, the degree of human attention. For example, if the edge distribution of a certain area spans a wide range and the connection with the surrounding area is also strong, it is likely that this area is on the boundary of a certain subject. The 16th circle is the internal structure circle \ of the division processing section 12 in the present embodiment. The other structure is the same as that of the first embodiment. The following description focuses on the parts that are different from the first embodiment. "In this circle, 29 is a hierarchy identification unit, 30 is a hierarchy parameter, and 31 is a division judgment unit. Figure 17 shows that 20 paper sizes are applicable to Chinese National Standard (CNS) Α4 specification (210X297 mm) (Please read the note on the back ^ n * before filling this page) • Α-χ *-装 ----- -Printed by A7 ________B7 printed by the Central Government Bureau of the Ministry of Economic Affairs and Consumer Cooperatives V. Description of the invention (18) The operation flow of the division processing unit 12 shown in Fig. 16 囷. As shown in Fig. 17 囷, the uniform division is performed first. (S26) e Next, the level parameter 30 of each area is determined in the level identification unit 29. (S27). The level identification unit 29 is based on the value of the divergence value α in the area and the state of the edge green distribution in the area β (including the edge green direction). , Distribution area, etc.), and the green connectivity γ to the edge of the surrounding area, determines the level parameter 30. For example, for the area where the divergence value α is smaller than a predetermined value, it is set to the lowest level (level C) ', and for the region with a larger value, the edge distribution β in the region is further obtained. The β value can be determined using the Sobel operator and the like described above. When the β value is smaller than the predetermined value, This area is considered a small area with independent, standing edges, rather than The boundary of the subject is set to a moderate level (level B). When the β value is relatively large, the connectivity γ is evaluated, and when the value of < / is large, it is classified into the most important level (level A) ° After hierarchical classification, 'calculate the motion parameter π in the motion calculation section 16', and the segmentation judgment section 31 first judges the critical value regarding the motion parameter (S28). Here, for the ^ field to be judged whether it should be divided In terms of 'the next step is to judge whether or not it should be divided based on the hierarchical parameter 30 (S29).' Then the division judgment section 30 determines in advance the extent of the area of each hierarchical level in which a judgment criterion for division is needed. When the hierarchical level is allowed When segmentation is performed, this region is segmented (S30). This is a process performed on all regions. Further, the same segmentation process is performed on all newly generated regions that are segmented (S33 to S38). In this embodiment, images are performed. When encoding, you can also take into account that this paper size is applicable to the Chinese national standard (CNS) A4 specification (210X297 mm) 17 ---------- 1 * (Please read the precautions on the back before filling this ) Printed by the Central Laboratories of the Ministry of Economic Affairs of the Bayer Consumer Cooperative Association *. A7 B7 V. Description of the invention (19) Image features across several areas, especially the contour line β of the subject. Therefore, it is possible to control In low areas, the amount of data is reduced by coarse coding, and at the same time, sufficient data is used in areas with low attention. [Fourth Embodiment] In the third embodiment, the level of human attention is used to determine the level. In this embodiment, the feature quantity of the image is used. In this embodiment, the consistency between the known feature quantity of the image and the feature quantity calculated in each area is used to store the known feature quantity. ) To determine the hierarchy. For example, there are quite a few studies on facial images so far, and various methods have been proposed to quantify the structure of the face as feature quantities. When these feature values are stored, a human face can be detected from the image (this is usually quite important). In addition, for other subjects, there are many situations that can be described by characteristic features based on brightness or texture data. When a human face is to be clearly expressed, for a region having a feature amount consistent with human features, it is set as the most important level A, and for other areas it is set as the level B of ordinary importance, and so on. Fig. 18 is a diagram showing the configuration of the hierarchical identification unit 29 in this embodiment. The other parts are the same as those in the third embodiment. In Fig. 18, 32 is a feature quantity memory difficulty, 33 is a feature quantity consistency calculation unit, and 34 is a hierarchy determination unit. In the feature quantity memory 32, it is applied to each paper size classified into a hierarchy. China National Chess Standard (CNS) A4 (210X297 mm) (Please read the notes on the back before filling out this page) — ^ 1 -I- H ^^ 1. ^ N ml «κι nl · rn v · ^ — HI 1 ^ 1 _, tr Printed by the Industrial Standards and Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 11 A7 B7 V. Description of the invention (20) Objects, storing the characteristics related to the objects. Consistency between the image 1 and the feature amount of the subject classified into each level. For example, the consistency may be obtained by obtaining an error between the feature quantity of the input image 1 and the feature quantity in the feature quantity memory 32. Next, the subject having the highest consistency is checked in the hierarchy decision section 34, and this region is classified into the hierarchy to which the subject belongs. According to this embodiment, the subject can be identified or checked out based on the feature amount of the image. And can improve the image quality of the required subject. Since the classification of the object level can be performed according to the feature quantity related to human attention, it can be encoded by taking into account human visual characteristics of the image. [Fifth embodiment] In the first embodiment, 'the distortion of encoding is taken into consideration in the integration process. In this embodiment, the encoding distortion is considered at the stage of the division process. Fig. 20 is a diagram showing the internal configuration of the division processing unit 12 in this embodiment. In this example, 35 is a division determination unit, and 36 is a division processing cycle instruction signal. The 21st circle shows the operation flow of the segmentation processing section 12 in FIG. 20. The segmentation processing section 12 of this embodiment uses the formula cited in the first embodiment. Using this formula, the initial segmentation processing is directed toward the frame. L (skn) is performed in a direction in which the total sum becomes smaller, which can reduce coding distortion when the same code amount is used. As shown in FIG. 21, firstly, the Chinese National Standard (CNS) Α4ϋΤΊϊ〇χ297 mm is applied to the uniform division 16 in the same paper size as the first stone paper. Arrvi-• · (Please read the precautions on the back before filling out this page),? Τ Printed by the Central Laboratories of the Ministry of Economic Affairs of the People's Republic of China, A7 ____B7_ V. Description of the invention (21) The state of the circle is uniform Block division (S39). This corresponds to the 0th division stage. The number of blocks obtained at this time is No, and each block is represented by B η. For each B%, it is judged whether to further divide the block. Compare L (BGn) related to BQn and the sum of L (SB0n (i)) related to each block obtained by dividing B 'into 4 blocks. If the latter is smaller, partitioning is allowed. For the calculation of the code size and distortion cost, first, BQn and SBGn (i) are coded in the temporary coding unit 22. Next, the decoding unit 23 generates a locally decoded video of BGn * SBGn⑴ from the encoded data obtained by the temporary encoding unit 22. Next, the encoding distortion calculation unit 24 calculates the distortion values D (BGn) and D (SBGn (i)) between the locally decoded image and the original image. The evaluation value calculation unit 25 calculates L (B〇n) and L (SB0n () based on the code amounts R (BGn) and R (SBGn (i)), the encoding distortion value D (BGn) * D (SB0n (i)), and i)) (S40, S41). The division judgment unit 35 compares L (BGn) with the sum of L (SB0n (i)) (i = 1, 2, 3, 4) of the four sub-blocks (S42). The BQn is divided into four SBGn (i) (S43). This corresponds to the first division stage. The block divided by SBGn (i) is represented by a new expression. For the same division judgment (S46 ~ S51), the same division process is performed a predetermined number of times. Finally, the division shown in Figure 8 is achieved. form. As described above, since calculation of the operation parameters is not performed in this embodiment, it is particularly advantageous in the case where a reduction in the calculation amount is required. [Sixth embodiment] Other embodiments are used to illustrate the paper size shown in Figure 9 in the first embodiment. The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm). (Please read the precautions on the back first. (Fill in this page again)

.1T Printed by Shellfish Consumer Cooperative, Central Standards Bureau, Ministry of Economic Affairs A7 B7 V. Description of Invention (22) Integration Processing Department 14. Fig. 22 is a diagram showing the internal configuration of the integration processing unit 14 in this embodiment. In this figure, 37 is a quantization parameter setting section, 38 is a quantization parameter, and 39 is a temporary encoding section. The operation of the integration processing unit 14 is basically the same as that of FIG. 10 except for S19. Chapter 23 is a flowchart showing the evaluation value calculation processing equivalent to S19. The calculation of the evaluation value is performed using the temporary encoding unit 39, the decoding unit 23, the encoding distortion calculation unit 24, and the evaluation value calculation unit 25. First, initial parameter values are set in the quantization parameter setting section 37 and output to the temporary encoding section 39 (S52). Next, the region Skn is coded in the temporary coding unit 39 (S53). When encoding is performed, quantization is performed using the set quantization parameter 6 so that the decoding unit 23 generates a locally decoded image from the encoded data thus obtained (S54). Then, the encoding distortion calculation unit 24 calculates a distortion value D (Skn) between the locally decoded image and the original image (S55). The evaluation value calculation unit 25 calculates L (Skn) based on the code amount R (Skn) and the encoding distortion value D (Skn) (S56). The cost value calculated initially is stored in Lmiji. After that, the quantization parameters are changed to perform the same cost calculation. Because the balance between the code size and the distortion value is changed by changing the Donghua parameter, the parameter with the smallest code size-distortion cost is used to finally set the code size-distortion cost L (Skn) of the area Skn (S57 ~ S60). After that, it is the same as that of the first embodiment. In this embodiment, the most appropriate integration process is realized in consideration of the quantization parameter. In addition, the method of adding Donghua parameters can also be applied to the code size-distortion cost-based segmentation described in the fifth embodiment. The paper size is applicable to China National Standard (CNS) A4 (210X 297 mm) -ml ^^ 1 —LI- HI 1 ^ 1 In n ** (Please read the note on the back before filling out this page) Order the seal of the Shellfish Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs A7 __B7 V. Description of the invention (23) Processing . [Seventh Embodiment] This embodiment is another example of the sixth embodiment. Fig. 24 is a diagram showing the internal configuration of the integration processing unit 14 in this embodiment. In this circle, '40 is a motion compensation prediction cost calculation section, 41 is a motion compensation prediction cost, and 42 is a temporary encoding section. The temporary encoding section 42 uses coding to determine motion parameters based on motion compensation prediction. At this time, the motion compensation prediction cost described in the first embodiment is used (Equation 2). That is to say, the determination of motion parameters during temporary encoding is performed at the least cost by using the balance point between the matching distortion caused by motion compensation and the code amount of the motion parameters. Specifically, the motion parameter is determined based on the cost value calculated by the motion-compensated prediction cost calculation unit 40 during the encoding process by the temporary encoding unit 42. Subsequent processing is the same as that of the sixth embodiment. In this embodiment, according to a predetermined constant λ, from motion compensation to coding, the total code amount_distortion cost can be minimized, and the shape of the area can be determined at the same time. As a result, the encoding distortion β can be reduced according to a predetermined code amount. [Eighth Embodiment] In this embodiment, an animation decoding device is described to encode encoding bits generated by using the various animation decoding devices described above. String for decoding. Fig. 25 is a configuration diagram of a decoding device. In this figure, '43 is the bit string analysis section, 44 is the area shape decoding section, 45 is the attribute information decoding section, 46 is the image data decoding section, and 47 is the motion information decoding «(Please read the precautions on the back before filling (This page). Binding. The size of the paper is in accordance with Chinese national standard (CNS > A4 size (210 × 297 mm). Printed by A7, B7, Consumer Work Cooperative of Central Procurement Bureau, Ministry of Economic Affairs. 5. Description of Invention (24) Department '48 Is the motion parameter, 49 is the motion compensation section, 50 is the predicted image, 51 is the image restoration section, 52 is the external memory boat, and 52 is the reproduced image. This decoding device displays the image frame or part of the image frame. Areas related to images (hereinafter referred to as "image frames, etc.") are divided into & state shape information, image data of each area encoded by a predetermined method, attribute information of each area, and motion information of each area Decode the encoded bit string, restore the regional image and reproduce the image frame, etc. In the case of this embodiment, 'because the encoding process is generated in the area outside the square' The description method of the area shape information is different from the conventional method. The description method used in this embodiment is: ⑴ It is represented by the vertex coordinates of each area; (Π) During the encoding, the area is divided and integrated. The process is represented. In the case of method (ii), for example, for any i and j ', the area number divided in the i-th division stage and the area number integrated in the j-th integration stage are described. The decoding device is the same as the encoding device. It is advanced after the "0th division stage of Fig. 6", and the final division state can be reproduced according to the same sequence of the decoding device. The amount of data in the case of method (ii) is generally It is less than the method of directly recording the coordinate data. The 26th circle indicates the operation flow of this decoding device. The encoded bit string 11 is first input to the bit string analysis section 43 so that the bit string is converted into encoded data (S61). From the encoding In the data, the area shape information is decoded by the area shape decoding unit 44, and the area of the image frame and the like is restored by the method described above. The paper size is applicable to China. Standard (CNS) A4 size (210X 297 public slime)

Ji —l · -rn 11 n I .. (Please read the note on the back before filling in this page) Order «Institute of the Central Ministry of Economic Affairs Bureau of the Bayer Consumer Cooperatives A7 _B7 V. Description of the invention (25) Cutting status ( S62). Using the restored area, the encoding order of the encoded area information in the subsequent bit string is specified. Each area is set to Sn. Then, according to the order of encoding, the data of each area is sequentially decoded from the bit string. First, the attribute information decoding section 45 decodes the attribute information of the area Sn, and then decodes the coding mode information of the area, etc. (S63). Here, if it is an intermediate mode (inter_mode), that is, when the prediction error signal is encoded (S64), the motion information decoding unit 47 decodes the motion parameter 48 (S65). The motion parameters 48 are then sent to the motion compensation section 49. Based on this, the motion compensation unit 49 calculates the desired memory address, which corresponds to the predicted image stored in the reference image stored in the external memory 52, and retrieves the predicted image 5 from the external memory 52 (S66) » The video data decoding unit 46 decodes video data of the area sn (S67). In the case of the intermediate mode ', using the addition of the decoded image data and the predicted image 50, a reproduced image of the final region center can be obtained. On the other hand, in the case of intra-mode (in-frame editing), the decoded image data is used as the reproduced image of the final area Sn. In order to use the reproduced image as a reference image for subsequent generation of the predicted image, the reproduced image is written into the external memory 52. As for these judgments and the principle of reproducing images, it is implemented in the image restoration unit 61 (S68). After all areas including video frames are implemented, this series of processing ends. The same process is performed for subsequent image frames and the like. 28 Guan Jiazheng Standard in this paper scale system ((: Gangba 4 Lai * (21 () / 297 mm) — I ------- Γτ ^ .-- •-(Please read the note on the back first Please fill in this page for further information), 1T printed by the Central Ministry of Economic Affairs of the Central Bureau of Zhuanzhou Bureau of Shellfish Consumer Cooperative A7 __B7_ V. Description of the invention (26) In the animation coding method of the present invention, not only the area is divided, but also the area. Integration, so that it can be edited flexibly in accordance with the structure of the image. As the judgment criterion for division or not is the result of comparing the encoding of the division when the region is divided and the division when it is not divided, "about the better The direction of encoding is exactly divided as required. Because the criterion for determining whether to integrate or not is based on the comparison of the advantages and disadvantages of the coding when the area is integrated and when it is not integrated, it can move in the direction of better encoding. On the other hand, the animation encoding device of the present invention includes a region division section and an encoding section, and the region division section includes a division processing section and an integration processing section. As a result, not only the region Segmentation also integrates regions, so it can be flexibly edited in accordance with the structure of the video. When the integration processing unit includes a temporary encoding unit, a decoding unit, a coding distortion calculation unit, and an evaluation value calculation unit, Under the condition of the predetermined code, the coding distortion is reduced. 'In the case where the segmentation processing unit uses the prediction error power as the motion parameter', it can further segment the region with large prediction error, that is, the region with large general motion. When the segmentation processing unit uses the edge strength of the original signal of each area as the motion parameter of the area, the shape of the area corresponding to the edge structure of the image can be obtained, and the subjective image quality such as the contour portion of the subject can be easily subject to The area of the image is further divided into 29 paper sizes that are in accordance with the Chinese National Standard (CNS) A4 specification (210X297 mm) (please read the precautions on the back before filling out this page). Industrial and Consumer Cooperatives Benben A7 ______ B7 V. Description of the Invention (27) — Use of the display area in the division processing section When the linear summation of the plural values of the image characteristics is used as the action parameter of the region, the region can be further divided according to several viewpoints or benchmarks. When the above plural values are also included in the motion of each region in the motion compensation prediction When the parameter code amount and prediction error power are used, the region can be segmented in a manner that reduces the total coding cost of the amount of information containing the motion complexity of the image and the amount of information of the motion parameter information, and can also be compared under the same distortion conditions. Encoding with a small amount of information "When the above-mentioned plurality of values can include the Ma Dong of the motion parameters of each region, the prediction error power in motion compensation, the divergence value of the original signal, the edge strength, and the size of the motion parameters of each region, Various reference values can be checked imaginatively to obtain the most appropriate area shape. In the case where the above-mentioned segmentation processing section includes a hierarchical identification section, it is possible to segment a relatively important region. In the case where the hierarchical recognition department observes the structure of the object across a plurality of areas, it is' easy to divide the area according to the shape of the object. When the structure of the object is based on the original signal of the area, the edge In the case of judging the intensity and the program connected to the surrounding area on the edge, the area shape can be obtained in accordance with the complexity of the original image signal or the structure of the object. In particular, it is possible to thin the area according to the outline structure of the subject. ”In the case of the feature quantity of the image observed by the hierarchical recognition unit, such as a person's face, etc., it is easy to divide the text according to the difference in details of each subject Paper standards are applicable to national standards for poor countries (CNS > A4 size (210X297 mm) (read the notes on the back before filling out this page), 11 Central Bureau of Standards, Ministry of Economic Affairs, Shellfish Consumer Cooperation, Du printed A7 B7 5 2. Description of the invention (28) Cut the area. Therefore, the shape of the area can be obtained according to the importance of the image area corresponding to the specific image and image characteristics of the person's degree of attention. When determining the level of each region by the consistency of the image feature quantity and the actual image feature quantity, the recognition rate of the object can be improved and the region can be segmented more accurately. The segmentation processing unit includes a temporary encoding unit, a decoding unit, and an encoding unit. In the case of the distortion calculation unit and the evaluation value calculation unit, the encoding distortion can be reduced under the predetermined code condition. The evaluation value calculation unit changes the quantization parameter in the same way. When the evaluation value is calculated, 'the optimization of the quantization parameter and the optimization of the area shape in the area coding can be realized at the same time, and the encoding efficiency is improved. When an evaluation value calculation section is provided in the previous section of the temporary encoding section, When calculating the linear sum of the code amount of motion parameters and prediction error power in each region in motion-compensated prediction, the motion parameters can be selected in a manner that minimizes the coding cost, and the region segmentation can be optimized. The method of reducing the total coding cost of the area including the optimization of the quantization parameter is to divide the area. On the other hand, the animation decoding device of the present invention has a region shape decoding unit and a video data decoding unit, so it can cope with Various shape areas generated in the encoding device. Therefore, it is easy to cooperate with the animation encoding device of the present invention. In the case where the area shape information includes relevant information about the process of segmenting and integrating each area during encoding, it is possible to use Reduce the amount of information to restore the shape of the area. This paper scale is applicable to the Chinese country 橾 隼(CNS > A4 specification (210X297 mm) ----------------, tr ------ Λ4 *. * (Please read the notes on the back before filling in (This page)

A simple explanation of the oxygen printing method by the Central Sample Rate Bureau of the Ministry of Economic Affairs of the Bayer Consumer Cooperative: Figure 1 is a common structural diagram of all the animation coding devices in the embodiment. Figure 2 is a flowchart of the operation of the coding device of the first frame. Fig. 3 is an internal configuration diagram of a region dividing section of a first circle. The fourth circle is an internal configuration diagram of the division processing section of the third ®. The fifth circle is a flowchart of the operation of the division processing unit of the fourth circle. Fig. 6 shows an example of the uniform division processing result in the division processing section of the fourth circle. Fig. 7 shows the results of the first initial segmentation processing in the segmentation processing section of the fourth circle. Fig. 8 shows the final result of the initial division processing in the division processing section in Fig. 4 ". Fig. 9 is an internal configuration diagram of the integration processing section in the third circle. The tenth circle is a flowchart showing the operation of the integration processing unit of the ninth circle. E FIG. 11 is a schematic diagram of an example of marking in the integration processing unit of FIG. Fig. 12 is a schematic diagram of an example of setting a Zhouyuan area in the integration processing section of Fig. 9. Fig. 13 shows the flow of step S19 in the tenth circle. "The fourteenth circle shows the internal structure of another embodiment of the segmentation processing unit in the third circle. Fig. 15 shows that in the division processing section of the 14th circle, the initial division of the paper size is applicable to the Chinese National Standard (CNS) A4 specification (2 丨 0 X 297 mm) ---.-------岿. (Please read the note f on the back before filling this page) Order the A7 B7 printed by the shellfish consumer cooperation department of the Central Bureau of Economic Affairs of the Ministry of Economic Affairs 5. The final result of the processing of the invention (30). The 16th round is the 3rd Figure 17 shows the internal structure of another embodiment of the segmentation processing unit. • The 17th circle shows the flowchart of the segmentation processing unit in Figure 16 during operation. The 18th figure shows the other embodiment of the 16th circle level recognition unit. 19th The figure is a schematic diagram of motion compensation prediction using the block matching method. Fig. 20 is an internal structure diagram of another embodiment of the segmentation processing section of Fig. 3. Fig. 21 shows the flowchart of the segmentation processing section of Fig. 20 during operation. Figure 22 shows the internal structure circle of the third embodiment of the integrated processing unit. Figure 23 shows the flowchart of the operation of the integrated processing unit at the 22nd cycle. Figure 24 shows the other processing unit of the integrated processing unit at Figure 3. [Internal structure diagram of the embodiment] FIG. 25 shows the embodiment The internal structure of the animation decoding device. Figure 26 shows the flowchart of the operation of the decoding device of the 22nd circle. Circle 27 is the schematic circle of the animation knitting device in the first known technology. Standard car (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before filling out this page)-Install. Order «Printing A7 B7 of the Bayer Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs of the People's Republic of China (31) Fig. 28 is a schematic diagram of an animation encoding device in the second conventional technique. Explanation of symbols: 1 ~ input image; 2 ~ area division; 3 ~ area shape information; 4 ~ area image signal; 5 ~ area motion information; 6 ~ Area attribute information; 7 to encoding section; 8 to locally decoded image; 9 to memory; 10 to reference image; 11 to encoded bit string; 12 to segmentation processing section; 13 to initial segmentation shape information; 14 to integration processing section 15 to uniform segmentation unit; 16 to motion calculation unit; 17 to action parameter; 18 to segmentation judgment unit; 19 to segmentation judgment instruction signal; 20 to marker unit; 2 surrounding area setting unit; 22 to temporary encoding unit; 23 ~ Decoding section; 2 4 ~ coding distortion calculation section; 25 ~ evaluation value calculation section; 26 ~ constant for calculating evaluation value; 27 ~ integration judgment section; 28 ~ integration processing cycle instruction signal; 29 ~ level identification section; 30 ~ level parameters; 31 ~ Segmentation judging section; 32 ~ Feature quantity memory; 33 ~ Feature consistency calculation section; 34 ~ Hierarchy determination section; 35 ~ Segmentation judgment section; 36 ~ Segmentation processing cycle instruction signal; 37 ~ Quantization parameter setting section; 38 ~ Quantization Parameters; 39 ~ temporary encoding section; 40 ~ motion compensation prediction cost calculation section; 4b motion compensation prediction cost; 42 ~ temporary encoding section; 43 ~ bit string analysis section; 44 ~ region shape decoding section; 45 ~ attribute information decoding 46 ~ image data decoding unit; 47 ~ motion data decoding unit; 48 ~ motion parameters; 49 ~ motion compensation unit; 50 ~ predicted image; 51 ~ image restoration unit; 5.2 ~ external memory; 53 ~ reproduced image; 101 ~ differentiator; 102 ~ prediction signal; 103 ~ prediction error signal; 104 ~ encoding unit; 105 ~ encoded data; 106 ~ decoding unit; 107 ~ encoded prediction error signal; 108 ~ adder; 109 ~ local solution Video signal; 110 ~ Memory; 111 ~ This paper size is applicable to Chinese National Standard (CNS) A4 specification (210 × 297 mm) (Please read the precautions on the back before filling this page). Packing. Order A7 B7 V. Invention Explanation (32) prediction section; 112 to motion vector; 113 to region division section; 114 to prediction section; 115 to region determination section; 116 to coding mode information; 117 to motion direction manager; 118 to coding section; 119 to coding data . (Please read the notes on the back before filling out this page) Install ‘

, 1T Printed by the Shellfish Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs This paper size applies to the Chinese National Standard (CNS) A4 specification (210 × 297 cm)

Claims (1)

The central government of the Ministry of Economic Affairs of the Central Government Bureau of Work and Consumer Cooperatives printed oxygen A8 B8 C8 ___D8 VI. Patent application scope 1. A method of animation coding, which is characterized by the following steps: According to a predetermined judgment criterion for segmentation, the input image is divided into multiple numbers. Area; for the divided plural areas, respectively, the area and its surrounding areas are integrated according to a predetermined integration judgment criterion, and the image signals are encoded for the areas left after the integration, respectively. 2. The animation encoding method as described in item 1 of the scope of patent application, wherein the above-mentioned judgment criterion for division or non-decision is determined based on the comparison of the coding pros and cons of a certain area and its area when it is divided and when it is not divided. 3. The animation coding method as described in item 1 or 2 of the patent scope of the application, wherein the above-mentioned criterion for determining whether to integrate or not is based on the compilation of a certain area when its area is harmonized with the surrounding area and when it is not integrated. The comparison of the pros and cons is determined. 4. An animation encoding device, comprising: a region division unit comprising a division processing unit for dividing an input image into a plurality of regions according to a predetermined division judgment criterion, and an integration processing unit for For the plurality of regions divided by the division processing section, the regions are integrated with their surrounding areas according to a predetermined integration judgment criterion; and the coding section is used for the regions left by the integration processing section. 'Encode their video signals separately. 5. The animation encoding device as described in item 4 of the scope of patent application, wherein the integrated processing unit includes: a male-time editing scorpion unit, which is used to pre-encode the image of each area and &lt; read the note on the back first (^^ items, then fill out this page) ------- ^ --- J ---- f I equipment -------- Order ----- ί-line -------- ---- Printed by ABC, B8, C8, D8 of the Central Cooperative Bureau of the Ministry of Economic Affairs 6. Apply for a patent and calculate the code size; The decoding department is used to decode the image encoded in the temporary encoding department; The distortion calculation section calculates encoding distortion using the image decoded by the decoding section, and the evaluation value calculation section 'calculates the evaluation value for judging the encoding quality based on the consideration of both the code director and the encoding distortion'; The comparison between the evaluation value obtained in the case of the surrounding area integration and the evaluation value obtained in the case of no integration determines whether or not to perform regional integration. 6. The animation encoding device according to item 4 of the scope of patent application, wherein the segmentation processing section includes: an action calculation section for calculating a prediction error power for judging the motion compensation prediction with each region as the motion parameter of the region; And a segmentation judgment unit for comparing the calculated motion parameter with a preset reference value; when the comparison result is that the motion parameter exceeds the reference value, _ is divided into a more detailed area. 7. The animation encoding device as described in item 4 of the scope of patent application, wherein the segmentation processing section includes: an action calculation section for calculating the edge strength of the original signal of each region as the motion parameter of the region; and segmentation judgment It is used to compare the calculated motion parameter with the preset reference value. When the comparison result is that the motion parameter exceeds the reference value, the paper size will be the Chinese standard (CNS) A4 (210X297) Mm) I. JJ <^ ~ binding / J ^ k (please read the back ¾. Note i before filling this page) Printed by the Central Standards Bureau of the Ministry of Economic Affairs, Shellfish Consumer Cooperative, A8, B8, C8, D8 6. Scope of patent application This area is divided into more detailed areas. 8. The animation encoding device according to item 4 of the scope of patent application, wherein the segmentation processing section includes: an action calculation section for calculating, for each region, a linear summation of complex numerical values representing the image characteristics of the region as the Motion parameters of the area; and a segmentation judgment unit to compare the calculated motion parameters with a preset reference value; when the comparison result is that the motion parameters exceed the reference value, this area is divided into more detailed areas " 9. The animation encoding device described in item 6 or 7 or 8 of the scope of the patent application, wherein the segmentation processing unit further includes a level identification unit for determining the importance of each region as a level parameter, using the action parameter and Importance to judge whether each area can be divided. 10. The animation coding device described in item 9 of the patent application park, wherein the hierarchy recognition unit observes the structure of the subject across a plurality of regions, and determines the hierarchy of each region. 11. The animation encoding device described in item 9 of the scope of patent application, wherein the hierarchical recognition unit observes the feature quantity of the image to detect the object, and determines the hierarchy of each area according to the result. 12. The animation encoding device according to item 4 of the patent application park, wherein the segmentation processing section includes: a temporary encoding section for preliminarily encoding an image of each area and calculating a code amount thereof; this paper scale is applicable to China National Standards (CNS &gt; Α4 specifications (210x297)) (please read the note on the back before filling in this page) Βϋ m HI —4— .nmm I ^ T—. Consumer cooperatives printed A8 B8 C8 D8 VI. Patent application range decoding section for decoding the image encoded in the temporary encoding section; encoding distortion calculation section that uses the image decoded by the decoding section to calculate encoding distortion; and evaluate Value calculation unit, which considers both the code size and encoding distortion. Next, it calculates the evaluation value used to judge the quality of the encoding; for each area, according to the evaluation value obtained when the area is divided into smaller areas and the undivided condition The result of the comparison of the evaluation values obtained below determines whether or not to divide the area. 13. The animation encoding device as described in item 5 or 12 of the scope of patent application, wherein in the temporary encoding section The quantization parameter of the prediction error signal accompanying the motion-compensated prediction is dynamically set, and the evaluation value calculation section calculates the evaluation value while changing the quantization parameter. 14. An animation decoding device that receives a coded image after it is divided into complex regions. The image encoding data includes: a region shape decoding unit for restoring the minus shape of each region divided during encoding according to the area shape information contained in the encoding data; and an image data decoding unit according to the restored data Each region shape confirms the encoding order of the region to decode the image of each region from the encoded data. 15. The animation decoding device described in item 14 of the scope of patent application, wherein the region shape information is included in the encoding, and each region is encoded. Information related to the process of segmentation and integration. Based on this information, the area shape decoding unit reproduces the same process as the encoding device to confirm the division status of the area. (Please read the note $ on the back before filling in this (Page)-Binding-Binding V-line The paper size is applicable to Chinese National Standard (CNS) A4 (210X297 mm)