TWI249907B - Method for fast mode decision of variable block size coding - Google Patents

Abstract

Methods for fast mode decision of variable size block coding referring to spatial and temporal correlations between a current encoding motion block and at least one reference motion block to decide a best mode for encoding the current encoding motion block. The at least one reference motion block includes at least one neighboring motion block of the current motion block and/or a previous motion block that is located in a previous image frame at a position corresponding to that of the current encoding motion block in a current image frame. At least one block size mode is obtained from the at least one reference motion block. The methods further check the reliability of the at least one block size mode before using the at least one block size to encode the current motion block.

Description

1249907 IX. Description of the Invention: [Technical Field] [01] The present invention relates to an encoding method for a variable size block, and more particularly to an optimal encoding mode for rapidly determining a variable size block of video image data Methods. [Prior Art] [02] To transmit multimedia data, especially dynamic video data, through a communication network, Lu needs to compress the data before transmission to match the available network bandwidth. Compression technologies such as MPEG-2, MPEG-4 and H.263 are currently technologies for compressing video data. The recently developed H.264 compression technology further improves the quality of compressed data. Compared with the previous shrinkage technology, the Η·264 technology can save time when the same compression quality is obtained. However, the amount of computation required by the 264.264 technology is much larger than the amount of computation required by previous compression techniques. [03] ITU-T's VCEG (Video) and the rib MPEG committee compared the H.264 rule. The Leg 4 technique includes seven block modes for inter-frame coding and two block modes for intra-frame coding. The two block modes used for intra-frame coding include the internal 6 mode and the handle mode. Each block has 8 _ directions to select the most suitable block mode according to the block's shank to improve compression efficiency. Further, for the _ touch, H264 technology provides multiple reference miscellaneous. From (4) reference frame cap correction - tilt the current frame most similar, test _ _. According to this method, the coding efficiency can be improved, and the accuracy of the displacement orientation prediction can be increased by 1/4 pixel. 1249907 [〇4] Although the H.264 technology has greatly improved the two causes: and more: Γ In fact, in the real one: "The current _ surgery becomes a challenge. Therefore, the need to simplify the coding cry _, so that H.264 technology in real time It is easier to apply in transmission. fInventive content] [〇5] Added _, the fast mode judgment method of the block size coding of this transmission day can save about _, 2: coding time required for the attack without sacrificing Excessive coding quality. In order to achieve the above object, the present invention discloses a variable-frequency touch method with a large variable block and a flat code, and it is known that Longda obtains at least a reference mode; The reference mode is executed = _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Each spring test plural reference dynamic block obtains the complex reference mode test mode, whether it can be estimated; and according to the dynamic evaluation result, each parameter is judged, if the fresh reference test shirt is the largest regional frequency coffee area Whether the difference vector of the displacement vector of the block is smaller than - ^249907 ^ I is difficult to cut (four) - difficult, It is determined that the transfer mode is reliable. For example, the displacement vector variation in the dynamic block can be determined by the displacement vector, and it is determined whether the displacement vector is greater than a second difficulty. Reference mode can be _ Γ 疋 [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ [ It is necessary to count the number of temples; the test ^1_ block adopts - the first - temple model, · the first - 杲 杲 仃 纽 纽 纽 纽 纽 · · 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果 如果Coding. 4, use the use of Tao Jin - step, when it is determined that there is no more than half _ complex reference kinetic energy region 'use the same first to help the heart [block mode for dynamic evaluation:; more than determined The reference mode of all the reference selections of the semi-test block is reliable; and the block is edited. 成本 Cost Co, the current dynamic zone details 3 features and implementations of the present invention, DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [Embodiment] A 3-square Yt is used for the block size mode of variable-size block coding. The common method requires _ _ (four) 彻 咖 TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM TM The heteroblock size mode is evaluated. Conversely, for the bottom-up merge method, the maximum block size mode is selected as the starting block size mode. The receiver's (four) method determines the starting block size mode according to the secret result. Whether the predetermined condition is met. If it is satisfied, the starting block size is modeled on the stone. Otherwise, the tilt block mode is selected to check the dynamic measurement, and the optimal block size mode is determined from the dynamic prediction result. Usually, in the case of recording a high bit rate, Lu's use a smaller block size mode to verify the opportunity, that is, to make the bottom up σ. In the case of lower bit rates, it is better to use a top-down separation method. Π6] can also make the best block size mode for other methods of secret coding, such as 'use medium size block mode as the starting area thief type to perform kinetic energy pre-^ followed by county, _ medium size block mode should be It is separated from other blocks I and commutate into a smaller block size mode.

The trick-method is to determine the displacement vector of a fine-block size pattern. The displacement vector prediction is a suitable block hit for encoding the entire motion_. The reading method analyzes the probability of the displacement vector to select a suitable block for the individual, 2 without reading each block, and the mode is performed. (4) Delete and then select - 5 are appropriate. The mode finance method can measure the size of various sub-blocks in a block. The lion money coffee __ rotten size model takes four displacement vectors. The displacement vectors of the other block size patterns are then predicted using these four displacement vectors 9 1249907. This method can only achieve dynamic prediction of displacement vectors that are distinct from the four nanovectors. [19] However, the above various methods do not consider the external information between the current dynamic block and its adjacent dynamic block, but only the sub-blocks in the current dynamic block. All the days, ?, the embodiment of the present invention, the fast mode judgment __ algorithm by determining the reference object, the secret block and the dynamic block of its adjacent dynamic block and the current dynamic block and the previous The temporal correlation of the reference dynamic blocks of the image frame determines the optimal block size mode for the doom prediction, wherein the previous image frame is located in the reference dynamic block corresponding to the current dynamic block in the current image frame. position. [20] Table 1 shows the experimental data of the spatial relationship between the current dynamic block and the adjacent scroll block in different encoders, such as foreman, COASTGUARD, CARPHONE, CONTAINER and ΑΚΓΥΟ. As shown in Table 1, the dynamic block adjacent to the current dynamic block is likely to be encoded using the same block size mode as the current dynamic block. Table 1 SPATIAL (2) SPATIAL (3) -——^ SPATIAL (4) FOREMAN 50·00ο/〇56.77% 63.58% COASTGUARD 54.17% 60.21% 65.17% CAJRPHONE 51.90% 59.08% *- 66.14% COOTAINER 79.46% 81.76% - --- 84.43% AKIYO 86.58% 88.43% 90.69% 10 1249907 [21] Table 1 shows that the current dynamic block and its adjacent sorrow block in different encoders make the 大_ block large dynamic Pre_ chance. Spatial (2) represents the adjacent dynamic block on the left and top of the target block (here, the adjacent dynamic block on the left and the top is the reference dynamic block) and uses any reference block. Block size mode to achieve the probability of dynamic prediction. Similarly, 趟^鲁(3)* SPATIAL(4) represents the probability that the current dynamic block uses any block size mode of its reference dynamic block to achieve dynamic prediction. In spatial(3), the reference moving block includes adjacent dynamic blocks on the left, upper and left sides of the ribbed block. In the SPA胤(4), the reference dynamic block further includes the right side of the current dynamic block. Adjacent dynamic block above. According to Table i, by referring to the block size mode of the adjacent dynamic block, the current dynamic block accurately predicts the area used for encoding: The probability of the size mode is about 60%. " • [22] In addition to the Lang, the present invention also considers the temporal relationship between adjacent image frames. In video images, the current video frame is usually very similar to its adjacent video frame. Therefore, dynamic blocks in two phase-inhibited frames, including static or low-dynamic targets or backgrounds, should be very similar, so the same block size can be encoded in 2 lines. That is to say, the dynamic block of the current video frame can use the block size mode of the dynamic block of the previous image = number as the prediction block size mode, and the first image frame in the first image frame is located in the corresponding image frame. The current location of the «Zone_Reference^ block. Currently before. w 11 1249907 ^ [ 3] Table 2 is not the result of the time interval of the dynamic block in the video frame using the previous image frame to test the block size mode of the heart block, where the parameter: dynamic block in The position in the previous image frame is in accordance with the position of the current frame of the current dynamic block.

[24] Table 3 further shows that after the reference lang and time correlation, the current dynamic block accurately predicts the probability of using the block size mode for encoding greater than 70%. Table 3 —--- -—.— TEMPORAL. SPATIAL (4) foreman ^ 75.51% COASTGUARD 77.46% CARPHONE ----- 77.26% CONTAINER '—— —— 89.98% AKIYO ^____ ^ 一—93.42% ~~-- 12 1249907 [25] FIGS. 1A and 1B are diagrams showing temporal and spatial information of a current dynamic block for obtaining a predicted block size pattern in accordance with an embodiment of the present invention. Figure 1A shows time information, wherein the current coded dynamic block]yfB is compared with a time reference dynamic block, wherein the position in the previous video frame (frame) of the reference dynamic block is in accordance with the current The dynamic block is located in the current image frame (Nth frame). Figure 1B shows spatial information, in which the current coded dynamic block touches the left adjacent neighboring block MB1 and the upper left adjacent dynamic block Mg. And the upper right side adjacent dynamic block MB3 is compared. [26] As mentioned above, the prior art Η264 technology analyzes seven possible block size patterns (ie 16x16, 16x8, 8x16, 8χ8, 8χ4, 4χ8 and CLASS) and selects one of the seven block size modes for Coding. On the other hand, the present invention only needs to analyze a subset of the seven block size patterns to reduce the coding time. According to the above, the new methodology can be used as a fast mode decision (FMD) algorithm. The FMD algorithm saves about 49% of the coding time. Then, the quality of the code is also reduced by about 0.6dB. ^ [27] In order to improve the coding quality of the above FMD algorithm, another real woman's J-step analysis The reliability of the prediction block size kernel is detected before encoding using the pre-hemp block size mode. If the prediction block size mode is reliable, the dynamic block will make the (four) Wei block size mode used for encoding. Otherwise, it will be throughout Dynamic block L-line Wang Hao mode search to find the best block size mode. This method is j 13 1249907 effective fast mode decision (EFMD) algorithm. [28] According to EFMD algorithm, FMD interpretation obtained Forecast It can be used for dynamic evaluation to obtain all possible pre-block large dumps. In order to judge the _ block size assignment, the job D algorithm considers the variation and difference value of the obtained displacement vector. If more than half of the predicted block size patterns are reliable, the FMD touch information can be considered reliable and can be used for encoding. [29] Figures 2A and 2B show large and small, respectively. The displacement is moving to the moving block. As shown in the figure, the -16χ16 dynamic block is divided into four% 8 blocks, that is, the block f is a bribe, the block _ one na 2, the block _ a fine, the district Block sub_MB4. In Figure 2A, each of the four 8χ8 blocks has a different displacement to 1, so that the entire dynamic block has a large displacement vector variation. In contrast, in the second diagram, four 8x8 Each block in the block has a similar displacement vector, so that the displacement vector variation of the dynamic block is small. In this case, the 16χ16 dynamic block itself can be used as the encoding. That is, when the displacement vector of the smaller block is used. When the variation is small, larger blocks surrounding smaller blocks are more suitable for coding. However, if the block size mode is 16x16, then there is only one set of displacement vectors after the dynamic evaluation. Therefore, the displacement vector variation cannot be calculated therefrom. Therefore, this embodiment only considers when the predicted block size mode is smaller than the 16x16 block. The displacement vector variation used to determine reliability in size mode. [30] The displacement vector variation can be calculated by the following equation. The block size pattern in which the displacement vector variation is greater than the variation threshold THvar can be determined according to the present invention 14 1249907 Reliable. = (MVx^VARCUT + MJ^y _VMcm) / 2 (1) MVx ^ VARcur = /1J; (MVxCUT - MVx:UT )2 (2) MVy — VAReur = ]^±{MVy^ ^ MVy^f (3) V nw=0 v 7 Lu MVxcur=(iMVCr)/n 9 MVycur^(^MVy^)/n (4) m=l MFxref = (gMVx::f)/n ^ MVy^ ^ };^ m=l , 'where η is the number of displacement vectors in each block, · χ and _ are the displacement vectors in the X direction and the y direction respectively, and the MVcur and MVref in the J knives are the current block and the reference area respectively. The displacement vector of the block. [31] Table 4 shows the statistical results of the reliable lure of the block size model for the displacement vector variation. As can be seen from Table 4, the average accuracy is greater than 85%. Table 4 P(B|A) P(C|A) FOREMAN 81% ----~- 19% COASTGUARD ----- 84% —---— 16% CARPHONE ------- 85% 15 % CONTAINER 92% 8% AKIYO 87% ! ---- ~~~ -———— 13% ——.— 15 1249907

Average g6% 14% A: MV VAR.^ > THvar B: Predicted block size mode correct C·Predicted block size mode error [32] As described above, only the ΐ6χΐ6 block size mode can be determined by the displacement vector variation. The reliability of the block size mode. For the 16χ16 block size mode, the reliability of the displacement vector can be judged from the difference value. If two adjacent blocks belong to the same target or have the same motion, then the probability that the two adjacent blocks are encoded using the same block size mode is high. Therefore, the displacement vectors of two adjacent blocks are also similar', that is, the difference values of the displacement vectors of two adjacent blocks are small. Conversely, if the displacement vectors of two adjacent blocks are different, it can be predicted that the two adjacent blocks have different motions, i.e., they do not use the same block size mode for encoding. Based on this idea, if the difference between the current block and the block adjacent to it is smaller than the difference threshold THmag, the block is considered to be reliable. The difference value Mag-difcur can be calculated by the following equation.

Mag^difcm =\MVxcur-MVxT^\MVycm-MVyJ (6) [3 3] Table 5 shows the reliable statistical results obtained by calculating the difference in displacement vectors. As can be seen from Table 5, the average accuracy is greater than 8〇0/. . Table 5 P(B|A) " ----~~-~_____ P(C|A) 75% ~------ 25% ' ---——

FOREMAN 16 1249907 COASTGUARD —.«««_ 70% —--- 30% CARPHONE —-—. —_ —------ 78% ___ 22% CONTAINER .....—------ ^ 93% —·~~~~-___ 7% AKIYO ""~TM" 1 ------ 94% ----__ 6% Average ----1 82% — 18% A : MH < Thmag B·Predicted block size mode is correct C·Predicted block size mode error [34] Figure 3 is a flow chart showing the process of judging sin by displacement vector variation and displacement vector difference . As shown in step 3G1, it is first determined whether the prediction block size (hereinafter referred to as the tea test mode) is the S block size mode. If 疋' is to perform the lamp step 3〇2, that is, the reliability of the reference mode is judged by the displacement vector difference value. If the result of the judgment is negative, that is, the reference mode is smaller than the application, step 303 is performed, i.e., the reliability of the predicted block size mode is judged by the displacement vector variation. [35] In step 3〇2, the difference value (Mag-difcur) of the displacement of the reference pattern can be calculated in accordance with equation (6) above. In step 3〇4, the difference value of the displacement Mag-dlfcur and the difference threshold ΤΗ· can be compared. As shown in step 3〇6, if Mag′′ is greater than THmag, the reference mode is considered to be “‘unreliable.” Otherwise, the reference mode is considered “reliable,” as indicated by the step tear. [36] In step 3〇3, since the reference mode is not the 16χ16 block size mode, ie, the smaller d block size mode, it is possible to compare the displacement vector variation if the displacement vector variation is the above-mentioned transformation (1) to the forest calculation bit 17 1249907 As shown, the reference mode can be considered as the reference mode as "movable vector variation MV-VARcur. In step 305, MV-VAR^ and the variation threshold THvar. As described above, MV-VARcUr is smaller than the variation threshold THyar, as in the step 3〇6 is “unreliable.” Otherwise, as shown in step 307, rely on “. [37] Through the above reference to Figure 2A and 2B ^ 绝民广奋士~曰 °9, in order to determine the best block size mode for the current coding block, the present invention ~~

The poor example refers to the current test block of the & block to obtain prediction information. Multiple block size patterns can be obtained by dividing and analyzing the prediction information. Usually, if some reference blocks are the same, the block size mode used by these reference blocks is also _. Further, if more than half of the reference blocks are off-block size mode, the probability that the current block is used to encode the same block size mode is very high. Table 6 provides experimental results for this hypothesis. < Table 6 P(B|A) P(C|A) FOREMAN 66% 34% COASTGUARD 62% 38% CARPHONE 70% ——— 30% ~—CONTAINER 93% ~——. 7% AKIYO —-—- _ 93% 7% Average --- --- · -- 77% ~*---- 23% —-— -----__ 18 1249907 A: More than half of the forecast information is the same B • Use large Most prediction information coding current block C does not use most prediction information coding. Most of the current blocks are as shown in Table 6. The current block uses the reference block obtained from its surrounding reference block to rise to 77%. Therefore, according to the present invention - Real method is 'if more than half _ reference phase _block size module: === time' then this method can judge that most of the fauna used is reliable if this majority is used The block is in the current block ιΓΠ Most of the blocks are used in the block size mode code search, in order to select the large dynamic block before the small block for all mode-style tearing her object judgment reference mode _ 中' 依考考娜驰_ The measurement information is obtained by yf] in the step operation, if more than half of the reference patterns are the same (ie, the same as the same day, step 403 is performed, otherwise, the step is performed. [42] in the step 4〇3, the disc estimate, in the step marriage, the inspection method can dynamically judge the reliability of the method described in Figure 3. The secret is good according to [] in ^4〇5, if this is the majority The reference mode used is 19 1249907, and step 413 is performed. Otherwise, all the modes are searched to select an optimal reference mode. _❿ block is the same, in step 4G7, when more than half of the reference modes are not Phase △ / Another half of the reference block uses the same - block large all-phase riding dynamic estimation, ... can check all the spring test mode. If the reliability is measured, he judges its reliability at the step. , ^ Take the method 1 according to Figure 3 [45] good 412 towel, _ material has a half: If not, as described in the step injury, the current dynamic search to select the best area Block size mode. Otherwise, the steps are executed; ^ Wang refers to the 'detection step _ to the best zone ghost size obtained by the step 核 no 疋 - δ χ 8 block size mode. If yes, as described in the step, continue for each 8 χ 8 blocks detect the best sub-partition. Since each _ block can be further divided into The 8χ4, 4χ8, and 4χ4 sub-blocks, the advantage of performing the step 在于 is that when it is detected that the dice block is not the best mode, it is not necessary to continue to analyze the sub-blocks smaller than the dice block, because the smaller sub-blocks The probability of becoming the best mode is very low. In this way, the coding time can be greatly reduced. 曰[47] Finally, in the still step, if the one-block size mode is not the most j-soil, it can be used to encode the current dynamic block. The mode with the lowest cost. [48] The fast mode decision method for variable block size coding disclosed in the above invention can greatly save the time of editing the dynamic block without compromising the code quality of the code 12 1249907. Table 7 and Table § show the experimental data of the results (ie, not = fast fast mode decision (Cong) algorithm search calculus tree face] listen _ all blocks

贝^果_ The surface algorithm of the present invention reduces the editing of the private mass by about q times and the continuation of the velocities (four) shape (four) paste (10). Under the condition of New Zealand _, the coding time of FMD is the beginning of the mouth. r-——---- Table 7 QP = 28 PSNR a difference foreman -0.10 dB COASTGUARD ——^ — -0.03 dB CARPHONE — -0.24 dB CONTAINER ---- -0.10 dB AKIYO -0.12 dB Average -0.12 dB

Bit rate difference 4.12% 2.47% 5.21 % 8.20 % 8.58 % 5.71 % Table 8 QP-28 Original "—·—^ FMD FOREMAN ~—. 47.3 sec ----— 29.18 sec COASTGUARD L— . 56.3 sec 30.73 sec 21 1249907 CARPHONE 42.01 sec 20.71 sec CONTAINER 40.63 sec 17.30 sec AKIYO 34.68 sec 16.19 sec Average 44.18 sec 22.82 sec [49] As shown in Table 9 and Table 10, the efficient FMD (EFMD) algorithm (ie, detection reliability) display ratio of the present invention is shown. The FMD algorithm has better results. The experimental results in Tables 9 and 10 show that encoding video images using the EFMD algorithm only reduces the coding quality by 0.2 dB, but saves approximately 44.3% of the coding time. Table 9 QP = 28 PSNR difference Bit rate difference FOREMAN -0.09 dB 1.16% COASTGUARD -0.04 dB 1.81 % CARPHONE -0.13 dB 1.38% CONTAINER -0.04 dB 2.98 % AKIYO -0.04 dB 1.42% Average -0.068 dB 1.75% Table 10 QP = 28 Original EFMD FOREMAN 47.3 sec 30.74 sec 22

Cheng Tu 1249907 COASTGUARD 56.3 sec 31.41 sec CARPHONE ----— 42.01 sec ~— 25.12 sec CONTAINER 40.63 sec 19.13 sec AKIYO 34.68 sec "~~ ~~------_ 17.72 sec ' ~ ~~-~- ~----- 24.82 sec Average 44.18 sec [50] Although the present invention has been described above in the preferred embodiment, it is not intended to be used in the present invention, and any skilled person skilled in the art, without departing from the spirit of the present invention, Within the scope of the consumption, the scope of the patent protection of the invention shall be subject to the definition of the patent application scope attached to the specification. BRIEF DESCRIPTION OF THE DRAWINGS FIGS. 1A and 1B are schematic views showing a time reference dynamic block and a four-year reference dynamic block of a current dynamic block reference according to an embodiment of the present invention, respectively.咏Θ and 2B are shown as schematic diagrams of reference dynamic blocks with large and small displacement vector variations. The figure is a flow chart for determining the reliability of the reference mode in accordance with the present invention, which is shown as a flow chart for determining the fast mode of the block size block according to the present invention. [Main component symbol description] 23

Claims (1)

1249907 X. Patent application scope: h - variable block size coding fast mode judgment method, including. From at least one reference _ block cap at least - Sensi mode; · Dynamic evaluation of the at least - reference mode According to the dynamic evaluation result, the block size mode is determined from the resident-reference mode towel to encode the current dynamic block. Take the soil 2·=Specialize 11 and tear the right (4) The rapid mode judgment of the marriage The V to the tea test dynamic block includes the current dynamic zone and one adjacent dynamic block. 3. If applying for the fast break method of the variable block size coding of the special fiber enclosure, the resident-reference dynamic_block includes the previous_block, = the current dynamic corresponding to the current video frame. The reference dynamic of the block is yjt» v〇y . The fast mode judging method of the Wei block size coding described in the second paragraph of the patent application, wherein the at least one adjacent dynamic block includes at least one dynamic zone located on the left side, the upper left side, the upper side or the upper right side of the current dynamic block Piece. A5. A fast mode decision method for variable block size coding as described in claim i of the patent application, wherein the towel further includes the step of detecting the reliability of the reference pattern. 6. The fast block mode break method of variable block size coding as described in claim 5, wherein the step of detecting the reliability comprises: the target dynamic block and the at least one reference dynamic block Performing a mobile bear estimation to obtain the displacement vector of the current dynamic block and the reference dynamic block, 24 However, the child's lesser test mode is not the maximum 1249907. If it is y $ test mode - the maximum block size mode, then compare. Determining a displacement vector difference value of the current dynamic block relative to the at least one reference dynamic block that produces the at least—reference mode; and determining the at least one reference mode if the displacement vector difference value is less than a first threshold For reliability. The method for determining the fast mode of the block size I is further included: and calculating a displacement vector variation of the current dynamic block; and if the displacement vector variation is greater than - Chu - After the threshold is reached, it is determined that the at least one reference mode is reliable. 8· a fast mode touch method for variable block size coding, including: obtaining a complex reference mode according to the complex reference dynamic block, performing dynamic evaluation for each reference mode; and judging each reference according to the dynamic evaluation result Whether the mode is reliable. For example, if you apply for the 8th job of the Secretary, the towel is more secret: Xiao Lian-style judgment is a reference _ block--parameter phase is small mode; "Maximum block size if the reference mode is Whether the difference value of the displacement vector of the maximum block large + block and the current dynamic block is small, determining that the reference is dynamic at a first threshold 25 1249907; and if the displacement vector difference value is less than the first threshold, Determining that the reference mode is reliable. The fast block of the variable block size coding described in claim 8 of the patent application is as follows: 'The method further includes' if the reference mode is not the maximum block size mode, the current dynamic zone is calculated. Block-displacement vector variation; whether the displacement vector variation of the _ secret block is greater than ~ threshold; and the reliability Γ the displacement vector variable is greater than the second value, determining the reference mode as the species variable region The fast mode judging method for block size coding includes: _ complex scale seams _ multiple reference patterns; formula; more than half of the complex reference dynamic block is adopted - first parameter Moss Lai-Mcco mode for reliability testing; and 4 vibrating fruit = tearing _, fine 帛, mode encoding at least - phase-faced thin accompaniment of the current dynamic block of the phase stability secret block and - the previous dynamic block, its In the position of the previous image frame 1249907, the current dynamic block is in the current frame 13 as in the patent application scope u. Judging method, among which, the fast mode of the stone horse is reliable 14目============================================================================================================ The reference mode is reliable; and the selection-reliable reference mode has the low cost of encoding the current _block. _ 15.1 The application of the fine _ 14 series of fast block size coding method, including If no more than half of the reference patterns are available (4), perform a full pattern search on the current dynamic region to find an optimal mode. 0, as described in Patent Procedural Item 11, the variable block size coding is fast. The speed mode judging method, wherein if the first reference block mode is a minimum block size mode, whether the block size pull can be further divided into smaller sub-block modes, the sub-control has a low cost coding The current dynamic block has a sub-block modulo 27 1249907. 17·===ΓΜ The fast mode of the variable block size coding described in item 4 includes whether the <minimum block size mode is guilty or appropriate, and if so, the minimum area is detected. The block size mode == continues to be divided by the sub-division zone, and the sub-block mode of the dynamic block is the lowest. 18·==Fast block size coding fast mode described in item 11 (4) Method 'The step of determining whether the first reference mode is reliable or not includes: Μ determining whether the first reference mode is a maximum block size mode If the tea reference method draws a large block size pattern, determine a displacement vector difference value between the meta-sorrow block of the first reference mode and the current block mode. If the displacement vector is different The value is less than a first value, and the determined mode is reliable; the J, the mode is determined to determine the reference mode. If the first reference mode is not the maximum block size, a displacement vector variation of the dynamic block, And if the displacement vector variation is greater than a second threshold, it is reliable. 28