TW462184B - Block motion video coding and decoding - Google Patents

Abstract

During transform-based video compression processing, motion vectors, which are identified during motion estimation and then used during motion-compensated inter-frame differencing, are constrained to coincide with block boundaries in the reference data. Block-based motion vectors have components that correspond to integer multiples of block dimensions. For example, for (8x8) blocks, allowable motion vector components are (..., -16, -8, 0, +8, +16, ...). Constraining motion vectors in this way enables the resulting encoded video bitstream to be further processed in the transform domain without having to apply inverse and forward transforms. In particular, an existing input bitstream is partially decoded to recover the motion vectors and prediction error (i.e., dequantized transform coefficients). Because the motion vectors coincide with block boundaries in the corresponding reference data, motion-compensated inter-frame addition can then performed in the transform domain to generate transform data for subsequent processing (which may ultimately involve re-encoding the transform data into another encoded video bitstream). Because motion compensation can be performed in the transform domain, the bitstream data can be further processed in the transform domain and without having to apply expensive and lossy inverse and forward transforms.

Description

V. Description of the invention (1) Background of the invention Field of the invention The present invention relates to algorithms for video compression and decompression. Cross Reference to Related Applications The present invention is filed for the following patent applications: US Patent Application No. 6 0/1 0 2, 2 1 4, 1 9 98 Application, File No .: SAR 1 2 5 8 5P; US patent application number: 60/1 21,531, filed February 25, 1999, file number SAR 1 299 0P. Related techniques are described in a typical transformation-based video compression algorithm, such as an algorithm that conforms to the Moving Picture Experts Group (MPEG) algorithm family, a block-based transformation, such as a separated cosine transformation ( DCT) is applied to the image data block corresponding to the pixel value or the pixel difference generated according to a motion compensation frame difference. After that, the conversion coefficients generated by each block are quantized in a general manner for subsequent encoding (for example: run-length encoding before variable length encoding) to produce an encoded video stream. According to a special video compression algorithm, images can be specified for compression processing of the following different types of data frames: ◦ An internal (I) data frame that is encoded using only the internal compression technology of the data frame, ◦ based on a correspondence Prediction data frame (P) that was coded using the data frame compression technique on the reference data frame of the previous I or P data frame, and the data frame itself can also be used to generate a reference data frame to encode more than one

6 2 18 4 V. Other information boxes of the description of the invention (2) 〇 Based on (i) the data frame pressure test comes from (i P data frame, or (for other data / idea: data can be used in p and B) The frame is decoded in any case, and the special pressure can be applied to the bit dequantization process to convert the applied data or (2) whether to use (1 set) separately. If addition is used, the shift data is used to generate the coded bit frame data during the solution. It can be seen that this step corresponds to the pre-application must be applied in many applications, as well as the forward and contraction technology) a iii)-the frame force mouth is pressed down inside the data frame to generate a flow of elements for the contraction process' To the solution pixel difference) The data is used for the motion compensation code pixel stream for decoding prediction. In the process of (H) reverse, or (lil) bidirectional prediction, the bidirectional (B) data frame is encoded. Three types of pre-I or P data frames, (ii) a successor or a combination, and the data frame itself cannot be used for coding. One or more image data reduction techniques are used for encoding. The resulting encoded video bit stream is reversed and inverted. For example, variable-length decoding ^, followed by run-length decoding, followed by destaring the transform coefficient block. After that, an inverse conversion coefficient block is used to generate (丨) decoded pixel heterogeneous data (based on the corresponding image data originally = internal or (2) data frame compression technology to compensate the data frame compression technology. The amount of motion compensation used in the analysis of the poor material frame is applied to the pixel difference. Ϊ 6 The amount of motion compensation in ί is determined by the compression process. Motion compensation reference material ^ a prediction of the material and pixel difference data ^. Β Therefore, Inter-frame addition of motion compensation data :: After the encoded video bit stream has been generated, traditionally, the encoded video bit stream has been previously

No? Page 4 6 2 18 4 V. Description of the invention (3) Generated by another (possibly remote) processor and considered as an input for the intended additional processing. For example, we may expect to perform the "transcoding" function, in which an existing encoded video bitstream (a bitstream that follows a video compression / decompression algorithm) is converted to a corresponding encoded video bitstream (Following a different stream of video compression / decompression algorithms). A "violent" method for implementing such transcoding operations is used to completely decode the input bit stream into the field of decoded pixels (based on the first video compression / decompression algorithm) and decode the resulting pixel data Fully encoded into the output bitstream (based on the second video compression / decompression algorithm). Another possible application is to insert a watermark into an existing encoded video bitstream. Similarly, a brute force method for watermark insertion is used to completely decode the input bit stream into the decoded pixel jaw field, and implement the decoded pixel processing in the pixel field to insert the desired watermark, and then modify it The subsequent pixel data is completely re-encoded to produce the desired processed output encoded video bit stream. In a traditional conversion-based video compression / decompression algorithm, the forward and reverse conversion steps can be quite expensive in terms of coding complexity and processing time. In addition, for missing conversions, the application of reverse conversion and subsequent re-application of forward conversion (during the processing of the violent volume of existing encoded video bitstreams) will generally result in Loss of information will result in degradation of decoded image quality (image quality of the reprocessed bitstream). Therefore, we look forward to being able to perform specific processing operations on an input bitstream (for example, with regard to transcoding or watermark insertion) without first having to fully decode the bitstream. If such processing is available at

4 6 2 18 4 V. Description of the invention (4) Performed in the field of transformation without causing a large amount of information leakage, can avoid expensive and leaky reverse and forward conversion steps. Summary of the Invention The present invention is applied to video compression and decompression technology, which can perform further processing on an existing scorpion video bitstream without first completely decoding the input bitstream. In particular, the present invention enables partially decoded video data to be processed in the field of conversion without generating a large (or possibly even any) loss of information contained in the input bit stream. Thus, the present invention avoids the need to apply expensive and leaky reverse and forward conversion steps to the processing of existing input bit streams. In accordance with the present invention, any mobile patcher video compression technique used to generate input coded video bitstreams is limited to motion vectors that are consistent with block boundaries. For example, in a station PEEG visual compression algorithm based on a (8X 8) DCT conversion, the 'movement vector element' is limited to an integer of eight seven (', the integer in can be positive, negative, or zero). Therefore, the processing operation can be applied to part of the decoded video data in the conversion field without generating a large amount of information loss (or possibly any). For a traditional mpeg video Λ, reduction / decompression algorithm, the input encoded video bit stream can be decoded with varying lengths, decoded by running length decoding, and decomposed to generate DCT coefficient data. Because the motion vector is limited to match the boundary of zone 1, afterwards, the interframe addition of motion compensation data can be performed in the DCT transformation field without first applying an inverse DCT transformation. After that, the movement compensation conversion data caused by mailing, registration, and / or conversion can be sub-jected to the special desired processing operation in the conversion field (for example, transcoding operation, watermark insertion

4 6 2 18 4 V. Description of the invention (5) Enter). After that, according to the application, the processed DCT coefficient data can be processed to at least some motion estimation, motion compensation data frame subtraction, requantization, run length recoding, and variable length recoding to produce the expected processing output. Encode the video bit stream without the need to implement separate reverse and forward DCT conversion steps. According to a specific embodiment, the present invention is a method for compressing video data. The method includes the following steps: (a) performing a motion estimation based on corresponding reference data in a data frame of a video data in order to identify a group of movements for the data frame Vector, where motion estimation is limited to identifying only block-based motion vectors that are consistent with the block boundaries in the reference and at least one of the block-based motion vectors is a non-zero motion vector (B) applying the inter-frame difference of the motion compensation data to the reference data and the block-based motion vector to the video data to generate the inter-frame difference data; and (c) one or more additional video compression steps It is applied to the data frame difference data to generate coding data for a coded video bit stream corresponding to the video data. According to another specific embodiment, the present invention is a method for processing an existing input encoded video bit stream. The method includes the following steps: (a) applying one or more decoding steps to the input bit Metastream responds to the conversion coefficient data in a transformation field and the corresponding block-based motion vector. The block-based motion vector is limited to the block boundary in the corresponding reference data. Consistent; (b) Performing an inter-frame addition of motion compensation data in the transformation field, generating prediction error correction (PEC) transformations based on block-based motion vectors and reference vectors

Page 10 4 6 2 18 4 V. Inventory May (6) coefficient data; and (c) follow-up processing of PEC conversion coefficient data in the conversion field. According to another specific embodiment, the present invention is a method for compressing video data, including the following steps: (a) performing a motion estimation in a data frame of a video data, and according to the corresponding reference data to form a data frame Identify a set of motion compensations, where the motion estimation is limited to identifying only block-based motion vectors (movement directions consistent with the block boundaries in the reference and all block-based motion vectors are (Zero motion vector); (b) applying the frame-to-frame difference of the motion compensation data to the video data, and generating the frame-to-frame difference data based on the difference data and the block-based motion vector so as to encode a video bit corresponding to the video data Metastream generates encoded data1 The encoded video bitstream follows a MPEG compression / decompression algorithm. The diagram briefly describes other aspects, features, and advantages of the present invention, which will be fully explained by the following detailed description, the patent application scope of the appendix, and the accompanying explanatory drawings, wherein: FIG. 1 shows a specific embodiment according to the present invention The obtained block diagram of the motion compensation data frame video compression processing; Figures 2 and 3 illustrate the traditional motion estimation processing (Figure 2) and the motion estimation processing of the (8 X 8) block of the present invention for image information (Figure 3) ) The difference between the two; FIG. 4 shows a block diagram of an in-frame video compression process for motion compensation data, according to an alternate embodiment of the present invention; and

Page 11 4 6 2 18 4 V. Invention (7) Fig. 5 shows a block diagram of a part of the decoding process, according to a specific embodiment of the present invention. DETAILED DESCRIPTION OF THE DRAWINGS FIG. 1 shows a block diagram of inter-frame video compression processing of motion compensation data, according to a specific embodiment of the present invention. According to this specific embodiment, the motion vector is restricted to coincide with the block boundary. In particular, the block-based motion estimation 102 is performed on input image data related to appropriate reference data. According to whether the data frame video compression processing is based on (i) forward prediction, (ii) backward prediction, or (i 1 1) bidirectional prediction, the reference data will be generated based on the following data: (1) a previous data frame , (I 1) —a subsequent data frame, or (1 1 i) a combination of the two (eg, the average of the previous and subsequent data frames). In any case, the motion estimation process attempts to identify a block of reference material (based on some similarity measures, such as: absolute pixel difference (S A D) that almost matches the current block of image data). According to the present invention, the block set of reference data used during the motion estimation process is limited to the block corresponding to the block limit. This limitation does not exist in traditional motion estimation processing, and the reference data blocks that can be used for motion estimation processing correspond to any pixel (or even sub-pixel) location within a specific search range. Figures 2 and 3 illustrate the difference between the traditional motion estimation process (Figure 2) and the motion estimation process (Figure 3) for (8 X 8) image data blocks of the present invention. Figures 2 and 3 show the search area of reference data for a special (8x8) image data block. The image data is displayed in a

Page 12 462184 V. Description of the invention (8) j (8 X 8) block (having a movement direction corresponding to the block center ㈢, 0) is shown in the figure f. The search areas in Figures 2 and 3 are based on moving inward. The size of each component of the motion vector is limited to 3 pixels (that is, the components of the motion vector can be varied between _ 8 and + 8 respectively. ). As shown in 囷 2, under the traditional integer-pixel-based motion estimation process, there are 28 9 different types available (8χ8). #The data box corresponds to 2 8 9 different integer-based motion vectors. The Motion vectors are based on a combination of 17 different integer component values (from _8 to +8). In the motion estimation scheme based on " broken pixels ", even more reference blocks are available. On the other hand, as shown in FIG. 3, for this particular case of the block-based motion estimation process of the present invention, the available reference data blocks are limited to 9 blocks corresponding to the block boundaries ( That is, it has one of 9 motion vectors obtained based on a combination of different component values (-8 ′ 0 ′ and +8). The traditional prediction based on this block-based motion estimation scheme will be more than the traditional pixel-based motion estimation scheme. Accurate, but the accuracy of block-based prediction is acceptable for multi-day applications. Since the number of available reference blocks has been significantly reduced for a specific search range (for example, the range in the example of Figures 2 and 3 is given by & 2 8 9 to 9) Block-based motion estimation processing is performed faster. Furthermore, the search range can be extended significantly > and it can reduce the processing time required for the traditional pixel-based motion estimation scheme. For example, extending the search range to ± 24 instead of ± 8 will still leave ^

Page 13 Case No. 88116755

4 6 2 Ί 8 4 Amendment V. Description of the invention (9) Only 4 9 (8 X 8) reference blocks for processing. In fact, the search range can be extended to ± 64, before the same number of reference blocks (ie, 289) can be used for pixel-based motion estimation (the search range is limited to ± 8). According to the special video compression algorithm and the shape of the data frame (for example, P or B), the block motion estimation process can identify three different block-based motion vectors (for each image data) Block): one based on forward prediction, another based on backward prediction, and the last based on bidirectional prediction. The mode control 104 process is performed after motion estimation to determine how to encode the current image data block, including the possibility of encoding using data frame internal coding techniques. According to the result of the control of the encoding mode, the corresponding motion compensation reference data may be used for the image data to execute the inter-frame difference 1 06 to generate block motion compensation data and the inter-frame pixel difference data. In any case, a transformation 1 08, such as an (8x8) DCT transformation, is applied to the pixel field data block to generate a block of conversion coefficients, and then the block is quantized 1 1 0 and run length 1 1 2 , And the processing of variable length coding 1 1 4 to generate a part of the coded video bit stream. Although not shown in Figure 1, at the appropriate time, the motion vectors used to encode the image data blocks are also encoded into the bitstream. Since the running length encoding 1 1 2 and the variable length encoding 1 1 4 are generally less missing encoding steps, one of the compression algorithms in FIG. 1. The decoding process (for example, the upper path in FIG. 1) can be started by The quantization coefficient data generated by the quantization block 11 0 will not affect the video compression / decompression

0: \ 60 \ 60523.ptc Page 14 2001.07.06.014 4 6 2 18 4 V. Description of the invention (10) 'I. — The danger of the reliability of the compression algorithm. In particular, the inter-frame addition of quantized motion compensation data from the conversion coefficients dequantized 6 and inverse transformation u 8 '2 / is used to generate the table ^ 日 日 ^ 机' test data or permission for another Group of video data bats / shells; people who are proficient in this skill will understand: In addition to the motion estimation of the present invention, the rest of the processing steps of Figure i: Ghost-based video compression technology ^ Standard: 2 to the traditional line. "敁 钿 Technology" and Figure 4 shows an inter-frame video diagram of motion compensation data. According to another embodiment of the present invention, the motion vector is shown in Figure 4. Restricted to AND region: two ;, the previous Γ 美 401 is applied to the motion estimation HOk-style controlled-side Si shift: this, block-constrained motion estimation 4.2, performed within the mode, And // compensation 4Q6 of the inter-frame difference is performed in the conversion field. From: t The specific embodiment of FIG. 1 is generally born in the pixel field. In the field of motion estimation and compensation is performed within the conversion field. The motion compensation data of the Z-beam path is added in the frame. 4 20 steps :: Execute within the π swap field. Therefore, no reverse conversion is required during the encoding process. The quantization step 4 1 0, the run length encoding 4 1 2, the variation length is again horse and the dequantization 4 1 6 can be the same as the similar steps of the specific embodiment of FIG. 8. FIG. 5 shows a block diagram of the binary division processing. 'According to the first aspect of the present invention, there are 2% of cases. The part of the decoding process of FIG. 5 is designed to view the current data in the existing, flat code (using the block-based motion compensation data of FIG. 1).

Page 15 V. Invention description "1) ------ = generated by the video compression algorithm). In particular, the partial solution j processes and receives an input bit stream and decodes it to produce a part. Transforming the conversion data ... 'Additional processing steps (for example: the step of inserting a watermark ::: watermark that is inserted and not shown in Figure 5) can be converted in: member 3 ”: to dequantize the conversion data to generate one After processing the conversion data, the two Qiao U Xuan Gong materials can be further encoded to generate an expected processed stream of Na-code video bits. Du Dijun, motion vector decoding 5 0 2 'mode decoding 5 04, and variable length : Hachiman length decoding / dequantization 5 06 is applied to the input

Gongxun (that is, do not do corpse 3 ^ "] pine-style control city J namely: interpolated ten,:, whether it is based on the river direction, backward, or both (also edited by M., A, 0, For example, the technique or the frame-space technique should be converted to the second domain ==? Conversion (for example,) coefficients (for A j- Θ, right is available, memory A 5 0 8 is reserved for the library in __诒 重 _, Conversion field reference material (for example: solution two:) ... ^ 1 ^ ^ t, 4 „#, ΓΛ Λ-Λ1'116 of the material frame, reference material (from memory. Flatten correspondingly; = = = = Don't get it) The interpolation reference data of the block measured by the average node and the four digits. It is expected that the state of the device 514 is determined by the decoding mode from block 504 (2; n: ated). If the type control The data pointed out | = That is, the automatic knife, 514 was switched to the clock to self-memory A will be suitable for apricot (also, motion compensation) follow-up data frame reference library into the plus 2 462184 V. Description of the invention (12).杲 The mode control data states "forward prediction," then the switch 5 i 4 is switched to a position where it can feed the appropriate previous data frame reference from memory B to the summing node 5 1 6. If The control data indicates "two-way prediction," then the switcher 5 is switched to a position where it can feed the appropriate interpolated tea test data from the average node 5 1 2 to the summing node 5 1 6. Finally, if the mode is controlled The data shows "internal-encoding", then the switcher 514 is placed on the ground to feed zero into the summing node 5 1 6. In any case, in the summing node 5 1 6 'from block 5 〇6's solution amount = D, CT coefficients are added by the switch 514 to the selected transformation-domain reference material to use the prediction error to correct the predicted value. Where appropriate (ie, as in 旲 a: 欠 田j I month / day · r node 51 6 conversion "material": box: 1 or? shell material box) 'from a total of 戍 more Λ skin feedback into memory Α Μ' for use in Reference materials that are encoded in boxes. For example, the DCT conversion is inside the line \ Limited to be consistent with the & block limit, and because the additions can be tested in the conversion, the total, node 5 1 6 motion compensation data is exchanged between frames. . Therefore 'except for the execution of 仃 within 1 without first applying a reverse turn 516 In addition to the migrating leakage and similar leakage, the conversion data of the total node hemp® & dynamic compensation corresponds to the conversion data that can be caused by Amu imku: image data. The: step is not shown in the figure: later, The additional code after processing is regarded as T; and oxime ^ s is used to generate the expected conversion collar, which is then transported to the next step: = 1. The external processing step can be performed, conversion collar calculation and motion compensation. In the appropriate; Motion estimation and / or motion compensation steps

P.17 462184 Five 'Explanation of the invention (13)'-*-*-f omitted. For example, the motion vector from the input bit stream is again used to make the second motion estimation impossible. If the output bit stream is generated without motion compensation, both motion estimation and motion compensation can be ignored. In addition to your compression / decompression algorithms, the present invention can also use the compression / decompression algorithms of M Field L, including algorithms based on conversions other than (8x8) and available for interpretation, and / or excluding operation. Length and the -crossing length encoding step and / $ 去 冬 古 # encoding step algorithm 3 has other additional post-quantitative editing. The present invention can be implemented in the way of Lei Luxiu, including including possible As the brain will understand: 2 people on the circuit element who are proficient in this skill are now a soft Zhao box 4 + = different functions can also be true in the digital field,-digital μ; 理! ^ 理step. This software can be used to 'exemplify m' as well as to deal with the problem of recognition & coin j as '5'. General-purpose electronic notebooks can be used in several ways to be used, have been used for many years. The present method of the device and other forms ~ a hard disk drive with a code embedded in the tangible medium, without removing the bar / *% The example of the medium is a floppy disk drive, CD-ROM, A try any other A storage medium that should contain r 渎 I taken from a fork, where 'the machine becomes hemp': eg, a computer) When it is loaded and executed, the two x-rays appear as a filamentous protrusion of the present invention (for example) ~ 4 kinds of storage media lament. The present invention can also be stored in a machine loaded by-machine and embodied in the form of: $ code 'should be used in some transmission wheels: 彳 γ or thin through fiber optic cable, or connected by cable) # 杀 "or Vehicles (such as excessive electrical wiring, "when the code is

Page 18

Case No. 88116755 Amendment 4 6 2 18 4 5. When the invention description (14 computers) is loaded and executed, the machine becomes a device for implementing the present invention. When executed on a general-purpose processor, the code section is combined with the processor to provide a unique device (which operates similarly to a specific logic circuit). I can still understand: the details, materials, and arrangements about the above descriptions and explanations to explain the essence of the present invention can also be made by those skilled in the art without departing from the following principles and scope of the patent application matters concerning the present invention Add various changes. Component symbol description 102 Block movement estimation 104 Mode control 106 Data frame difference 108 Transformation 110 Quantization 1 12 Run length 114 Variable length coding 116 Dequantization 118 Inverse transformation 120 Motion compensation data inter-frame addition 40 1 Transformation 402 Block-constrained motion estimation 404 mode control 406 inter-frame difference of motion compensation data 410 quantization 412 run length coding 414 variable length coding Shima 416 dequantization 420 motion compensation data inter-frame addition 502 motion vector decoding 504 pattern decoding 506 variable length decoding // run length decoding / solution Quantization 5 08 dequantization conversion system [memory A 5 10 dequantization conversion system memory 1 memory B 5 12 average hfz point 5 14 four-position switcher 516 plus i'k point

0: \ 60 \ 60523.ptc Page 19 2001.07. 06.019

Claims (1)

Case No. 881 16755 6. Scope of patent application | cy ^ year " 7 ear worship! &Quot; ΤΤΤΤ ^ ζ ^ Γ correction i. A method for compressing video data, including the following steps: Ca) According to the corresponding reference materials Execute the video data t motion estimation of a data frame to identify a set of motion vectors for the data frame, where the motion estimation is limited to identifying only block-based motion vectors, which are related to blocks of reference data Consistent boundaries, and at least one block-based motion vector is a non-zero motion vector; (b) applying the reference frame and the block-based motion vectors to apply the inter-frame difference of the motion compensation data to The video data to generate data frame difference data: and (c) apply one or more additional video compression steps to the data frame difference data to generate coded data for a coded video bit stream corresponding to the video data . 2. The method according to item 1 of the scope of patent application, wherein steps (a) and (b) are performed in a pixel field and step (c) includes the following steps: (1) applying a transformation to the data frame difference Data to generate conversion coefficient data in a conversion field; and (2) quantize the conversion coefficient data to generate quantized conversion coefficient data. 3. The method according to item 2 of the patent application, wherein the encoded video bit stream follows an MPPEG compression / decompression algorithm, and the conversion is a discrete cosine transform (DCT). 4. The method according to item 1 of the patent application scope, wherein the Yin step (a) further includes the step of applying a conversion to the pixel data 1 to generate a conversion coefficient data in the conversion data, wherein the motion estimation is based on a reference in the conversion field. The data is applied to the conversion coefficient data. 6052J.p [c Page 1 2001.09. 05.022 4 6 2 1 8 4 Case No. 88116755 ----------------------- ------- ------- # κ "Revision 6" Application for patent scope yu 丨 5. A machine-readable medium has encoded code 1 on it 1 of which 1 when the code is executed by a machine The machine performs the following steps; (a) performs the motion estimation based on the corresponding reference data in the j-video frames to identify a set of motion vectors for the data frame, where the motion estimation is limited to identifying only the Block-based motion vectors with consistent block boundaries and at least—the block-based motion vector is a non-zero motion vector> (b) The inter-frame differences in motion compensation data are based on the reference and Block-based motion vectors are applied to the video material to make Generate data frame difference data and (c) apply one or more additional video compression steps to the data frame difference data ϊ in order to generate coded data for an encoded video bit stream corresponding to the video material 0 6 types Method for processing an existing input 'encoded video bitstream The method includes the following steps (a) Applying-^ or more decoding steps to the input bitstream in order to transform fields and corresponding blocks in The conversion coefficient data is returned in the motion vector based on which the block-based motion vector is limited to correspond to the block limit in the corresponding reference material-so (b) in the conversion field based on the block-based The motion vector and the reference data are used to perform an inter-frame addition of the motion compensation data > to generate a prediction error correction (EPC). Data; and (c) after implementation on the EPC conversion coefficient data in the conversion domain 60523 -ρΐ c Page 2 2001.09.05. 023 4 6 2 18 4 Amendment No. 881 16755 VI. Scope of Patent Application Continued processing. 7. The method of claim 6 in which at least one block-based motion vector is a non-zero motion vector. 8. If the method according to item 6 of the patent application is applied, the coded video bit stream conforms to a MPEG encoding and decoding algorithm, and the conversion field corresponds to a DC conversion field. 9. The method according to item 6 of the patent application, wherein step (c) comprises applying one or more video compression steps to generate a processed output encoded video bit stream. 10. The method according to item 9 of the scope of patent application, which causes one or more video compression steps to include the motion compensation data frame difference to the processed conversion coefficients in the conversion field to generate conversion coefficient difference data. 1 1. A machine-readable medium having encoded code thereon, wherein when the code is executed by a machine, the machine completes the following steps: (a) applying one or more decoding steps To an existing input video bit stream 1 to restore the conversion coefficient data in a conversion field and the corresponding block-based motion vector, where the block-based motion vector is limited to the corresponding The block boundaries in the reference material are consistent; (b) In the conversion field, an inter-frame addition of motion compensation data is performed based on the block-based motion vectors and the reference material to generate a prediction error correction (PEC) conversion. Coefficient data; and (c) in this conversion area, after implementation on P EC conversion coefficient data 6052j.ptc Page 3 2001.09.05.024 Case 婼 88116755 Scope of Patent Application Continued processing. 12. One (a) motion estimation 1 motion vector, and all block-based inter-frame difference (c) difference data stream generated decoded algorithm based on which, the block will shift the basic data will be one, using the encoding method. The video information corresponds to the movement direction of the motion compensation data based on the motion estimation motion vector; and the method of one or more correction materials, including the following steps: Data, the execution of the movement meter on a video data frame is limited to identifying only The block-based basis is consistent with the block boundaries in the reference material, and the motion vectors are all zero motion vectors. The inter-frame difference is based on the reference material and the amount is applied to the video data to generate additional video compression steps. When applied to the data frame, it is an encoded video bit data corresponding to the video data, where the encoded video bit stream conforms to a MPEG coding 6052J.ptc Page 4 2001.09.05.025