CN101931819B - Temporal error concealment method - Google Patents

Abstract

The invention provides a temporal error concealment method, which comprises the following steps of: detecting a damaged macro block which consists of four 8*8 damaged blocks; acquiring motion vectors of multiple 4*4 adjacent blocks encircling the damaged macro block; and aiming at each 8*8 damaged block, determining a predicted motion vector of the 8*8 damaged block according to the motion vectors of six 4*4 adjacent blocks closest to the 8*8 damaged block in the multiple 4*4 adjacent blocks.

Description

Temporal error concealment method

Technical field

The invention relates to a kind of temporal error concealment method, particularly can predict motion-vector, cutting apart pattern and using partial distortion (partial distortion) comparison method to carry out the motion-vector procedure time property error concealing method of refining according to adjacent block based on the motion-vector predicted decision block about a kind of.

Background technology

Along with application of multimedia technology is more and more welcome,, has developed and many video coding techniques for effective compressed video file.The purpose of compression is in order to remove the data of unnecessary (redundancy) in the view data, with the storage area of reduction image or the transmission quantity of image.In compression standard H.264, intra-frame prediction (intra prediction) and two kinds of predictive codings of inter-picture prediction (inter prediction) have been adopted.Intra-frame prediction is to utilize in the same picture adjacent block correlation spatially to predict, inter-picture prediction then is to utilize between adjacent pictures correlation in time to predict.

To inter-picture prediction, H.264 compression standard has defined the pattern of cutting apart of 7 kinds of different blocks sizes to each 16 * 16 macro zone block: 16 * 16 (T1 patterns), 16 * 8 (T2 patterns), 8 * 16 (T3 patterns), 8 * 8 (T4 patterns), 8 * 4 (T5 patterns), 4 * 8 (T6 patterns), 4 * 4 (T7 patterns), as shown in Figure 1.Behind the coding, each block can be represented by residual value (residual) and motion-vector (motion vector).Select the more little block coding of size, can obtain good more image quality, but required operand and time are also big more.Therefore, in order to take into account image quality and code efficiency, generally be different and the blocks different sizes of employing, to obtain more to compress usefulness according to picture complexity.

Video data becomes after overcompression and is easy to the bit stream (bitstreams) that transmits and store; Yet the video bit stream of these high compression is easy to take place for example packet loss problems such as (packet erasure) in the process of transmission (especially under radio channel environment).For fear of because of the video packet loss influences the video pictures quality, protection mechanism commonly used has automatic repeat requests (ARQ), forward error correction (FEC) and error concealing (Error Concealment).Compared to ARQ and FEC, Error concealment techniques does not need extra frequency range, and under the situation of broadcasting (Broadcast) and multiple transmission (Multicast), preferable usefulness is arranged.Mainly be divided into two kinds in the performed Error concealment techniques of decoding end: spatial error concealing (spatial error concealment) and timeliness error concealing (temporal errorconcealment); Wherein mainly to utilize spatial redundancy information in the same picture to come restoring damaged video sequence, temporal error concealment then be to utilize in the coded sequence high correlation between each continuous pictures to rebuild impaired sequences to spatial error concealing.Owing to generally all exist very high correlation between adjacent pictures; Therefore compared to spatial error concealing; Except some special circumstances (occur suddenly like scene conversion, object or disappearance etc.), temporal error concealment method can provide preferable image quality usually.

When the intersexuality error concealing method is rebuild impaired block in use,, therefore at first need obtain the motion-vector of impaired block owing to will utilize the information of other reference picture.Developed at present and the motion-vector of some simple methods in order to the predictive of impaired block, for example measurable motion-vector is 0, on the usage space motion-vector of adjacent block mean value or use motion-vector or the like the method for same position block in the reference picture.Said method can with reference to by people such as Y.Wang in Proc.IEEE, vol.86, no.5; Pp.974-997; " the Error control and concealment for videocommunication:a review " that is delivered among the May 1998, by people such as D.Agrafiotis in IEEE Trans.Circuits Syst.Video Technology, vol.16, no.8; Pp.960-973; " the Enhancederror concealment with mode selection " that is delivered among the Aug.2006, by people such as S.Valente in IEEE Trans.Consumer Electronics, vol.47, no.3; Pp.568-578; " the Anefficienct error concealment implementation for MPEG-4 video streams " that is delivered among the Aug.2001, by people such as B.Yan in IEEE Trans.Consumer Electronics, vol.49, no.4; Pp.1416-1423; " the A novel selective motion vector matching algorithm forerror concealment in MPEG-4 video transmission over error-prone channels " that is delivered among the Nov.2003, by people such as J.Zhang in IEEE Trans.Circuits Syst.Video Technol., vol.10, no.4; Pp.659-665; " the A cell-loss concealment technique for MPEG-2coded video " that is delivered among the Jun.2000, by people such as J.Y.Pyun in IEEE Trans.Consum.Electron., vol.49, no.4; Pp.1013-1019; " the Robust error concealment for visualcommunications in burst-packet-loss networks " that is delivered among the Nov.2003 and by people such as S.C.Huang in inProc.Int.Conf.MultiMedia Modeling (MMM), Jan.2008, LNCS 4903; " Temporal Error Concealment for is the Using OptimumRegression Plane H.264 " that is delivered among the pp.402-412, content will be incorporated this paper into as a reference on it.

In addition, other various modification methods that proposed to temporal error concealment technology, also can with reference to by people such as M.E.Al-Mualla in Electron.Lett.; Vol.35, pp.215-217, " the Temporal error concealment using motion field interpolation " that is delivered in 1999, by people such as S.Tsekeridou in IEEE Trans.Multimedia; Vol.6, no.6, pp.876-885; " the Vector rational interpolation schemes for erroneous motion field estimationapplied to MPEG-2 error concealment " that is delivered among the Dec.2004, by people such as J.Zheng in IEEE Trans.Broadcast., vol.49, no.4; Pp.383-389, " the A motion vectorrecovery algorithm for digital video using Lagrange interpolation " that is delivered among the Dec.2003, by people such as J.Zheng in IEEE Trans.Multimedia, vol.6; No.6; Pp.801-805, " Error-concealment algorithm for is the using first-order plane estimation H.26L " that is delivered among the Dec.2004, by people such as J.Zheng in IEEE Trans.Multimedia, vol.7; No.3; Pp.507-513, " Efficient motion vector recovery algorithm for is the based on apolynomial model H.264 " that is delivered among the Jun.2005, by people such as S.Shirani in IEEE Journal on Selected Areas inCommunication, Vol.18; Pp.1122-1128; " the A concealmentmethod for video communications in an error-prone environment " that is delivered among the June 2000, by people such as Y.C.Lee in IEEE Trans.Image Process., vol.11, no.11; Pp.1314-1331; " the Multiframe error concealment for MPEG-coded video delivery overerror-prone networks " that is delivered among the Nov 2002, by people such as G.S.Yu in IEEE Trans.Circuits Syst.VideoTechnol., vol.8, pp.422-434; " the POCS-based errorconcealment for packet video using multiframe overlap information " that is delivered among the Aug.1998, by people such as S.Belfiore in IEEE Trans.Multimedia; Vol.7, no.2, pp.316-329; " the Concealment of whole-frame losses for wireless low bit-rate videobased on multiframe optical flow estimation " that is delivered among the Apr.2005, by people such as P.Baccichet in IEEE Trans.Consumer Electronics; Vol.51, no.1, pp.227-233; " Frameconcealment for is the decoders H.264/AVC " that is delivered among the Feb.2005, content will be incorporated this paper into as a reference on it.

Though existing many researchs improve to temporal error concealment method, the forecasting accuracy of motion-vector and the effect of compensation and efficient still have the space of improvement.

Summary of the invention

In view of the existing problem of prior art, the invention provides a kind of high-effect temporal error concealment method that is applicable to H.264, can improve the forecasting accuracy and the error concealing usefulness of motion-vector effectively.

According to an aspect of the present invention, a kind of method that is applied to the temporal error concealment in the video decode is provided, the method includes the steps of: a. detects an impaired macro zone block, and this impaired macro zone block is made up of 48 * 8 impaired blocks; B. obtain motion-vector around a plurality of 4 * 4 adjacent block of this impaired macro zone block; C. to each this 8 * 8 impaired block, by in these a plurality of 4 * 4 adjacent block near the motion-vector of 64 * 4 adjacent block of this 8 * 8 impaired block, determine a prediction motion-vector of this 8 * 8 impaired block; D. this prediction motion-vector of these 48 * 8 impaired blocks is relatively merged into 2 16 * 8 in twos with decision nonjoinder these 48 * 8 impaired blocks, with these 48 * 8 impaired blocks and is merged impaired block, these 48 * 8 impaired blocks are merged into 28 * 16 in twos merges impaired block, maybe these 48 * 8 impaired blocks all merged into 1 16 * 16 and merge impaired block; E. to after merging this 16 * 8 merge impaired block, this 8 * 16 merges impaired block or this 16 * 16 and merges impaired block, determines one of the impaired block of each this merging to predict that motion-vector merges the mean value of the prediction motion-vector of 8 * 8 impaired blocks that impaired block comprises for this; And f. is to each this 8 * 8 impaired block, the impaired block of this 16 * 8 merging, this 8 * 16 impaired block of merging or the impaired block of this 16 * 16 merging; With this impaired block respectively or to merge pairing this prediction motion-vector of impaired block be a starting point; Carry out the pixel comparison with a plurality of reference block in the search window in the reference picture, in this search window, to seek and this impaired block corresponding reference block of being complementary of the impaired block of this merging maybe.

Others of the present invention, part will be stated in follow-up explanation, and part can be learnt in explaining easily, or can be learnt by embodiments of the invention.Each side of the present invention is with the element that is particularly pointed out in the above-mentioned claim capable of using and combination and understand and reach.Need to understand, usefulness is for example all only made in aforesaid summary of the invention and following detailed description, is not in order to restriction the present invention.

Description of drawings

Graphic is to combine with this specification and constitute its part, in order to the explanation embodiments of the invention, and together with specification in order to explain principle of the present invention.Said embodiment is preferred embodiment of the present invention, yet, must understand configuration and the element of the present invention shown in being not limited to, wherein:

Fig. 1 is shown as the pattern of cutting apart to the defined different blocks size of inter-picture prediction of compression standard H.264;

Fig. 2 is the flow chart of temporal error concealment method provided by the present invention;

Fig. 3 illustrates in the process of transmission and loses or an impaired macro zone block, and contiguous a plurality of 4 * 4 blocks;

Fig. 4 is upper left 8 * 8 blocks of macro zone block among Fig. 3;

Fig. 5 illustrates the active plane of regression that is calculated according to 64 * 4 adjacent blocks in one embodiment;

Fig. 6 shows the method flow diagram in order to decision variable region block size pattern that one embodiment of the invention provides;

The comparison order that the one embodiment of the invention that illustrates Fig. 7 adopts; And

Fig. 8 and Fig. 9 are for T4 (8 * 8) pattern being the sketch map and the flow chart of the program of refining of example explanation one embodiment of the invention.

[main element label declaration]

300 impaired macro zone blocks

300A, 300B, 300C, 300D 8 * 8 blocks

310,312,314,316 4 * 4 blocks

320,322,324,326 4 * 4 blocks

330,332,334,336 4 * 4 blocks

340,342,344,346 4 * 4 blocks

500 active planes of regression

8008 * 8 impaired blocks

810,8208 * 8 adjacent block

812,814,816,8184 * 4 blocks

822,824,826,8284 * 4 blocks

Embodiment

In order to make narration of the present invention more detailed and complete, can and cooperate that Fig. 2's to Fig. 9 is graphic with reference to following description.Device, element and method step described in right following examples in order to explanation the present invention, is not in order to limit scope of the present invention only.

Fig. 2 is the flow chart of temporal error concealment method provided by the present invention.At first, at step S200, reception and decoding comprise a vision signal of a plurality of pictures (frames), and wherein at least one picture has to be lost or impaired macro zone block.In step S210,, determine the motion-vector of adjacent block on its space to a loss or an impaired macro zone block.Then, in step S220, a plane of regression (active regression plane) and dope the motion-vector of impaired block initiatively that uses that motion-vector institute construction by adjacent block goes out.In step S230,, determine the pattern of cutting apart of impaired macro zone block according to the motion-vector of the impaired block that is doped.At last, in step S240, use the motion compensating method of variable resource block size to carry out refine (the motion refinement) of motion-vector, in reference picture, to hunt out better compensation block.In addition, but the present invention proposes the method for several premature termination steps S240, with the minimizing program required time of refining, promotes compensation efficient.Below will do more detailed description to step S220 to S240.

With reference to figure 3; Macro zone block (MB) 300 is loss or one 16 * 16 impaired macro zone blocks in the process of transmission, and block 310,312,314,316,320,322,324,326,330,332,334,336,340,342,344 and 346 is vicinity 4 * 4 blocks of MB 300.The present invention utilizes the high correlation between the motion-vector of block adjacent one another are on the space to come the motion-vector of predictive of impaired MB 300.At first, as shown in Figure 3, impaired MB 300 is divided into 48 * 8 blocks (upper left block 300A, a left side are descended block 300B, bottom right block 300C, reached upper right block 300D).The present invention is according to the correlation between each impaired 8 * 8 block 64 * 4 blocks adjacent thereto; And propose one initiatively plane of regression predict the motion-vector of each impaired 8 * 8 block, the second order plane that this plane calculates for position and corresponding motion-vector thereof according to adjacent block.

With reference to figure 4, be that example is explained the method for the present invention with active plane of regression prediction motion-vector with upper left 8 * 8 block 300A among Fig. 3.Because more approaching block each other generally has high more correlation on the space, so selection of the present invention and hithermost 6 blocks 312,314,316,320,322,324 of block 300A come the motion-vector of calculation block 300A.The center point coordinate of block 300A is set at (0,0), and the center point coordinate of then adjacent 64 * 4 blocks 312,314,316,320,322,324 is respectively (6; 6), (2,6), (2,6), (6; 2), (6;-2), and (6 ,-6), and its corresponding motion-vector is expressed as V1, V2, V3, V4, V5, and V6 respectively.Active plane of regression proposed by the invention is:

Z _i(x，y)＝α ₁x _i ²+α ₂?x _i?y _i+α ₃y _i ²+α ₄x _i+α ₅y _i+α ₆

Wherein x and y represent the center point coordinate of adjacent 64 * 4 blocks, and z is pairing motion-vector.With the centre coordinate of adjacent 64 * 4 blocks and corresponding motion-vector V1, V2, V3, V4, V5, and V6 bring following formula into, can calculate alpha ₁, α ₂, α ₃, α ₄, α ₅, and α ₆With reference to figure 5, it illustrates the active plane of regression 500 that is calculated according to 64 * 4 adjacent blocks in one embodiment, and the motion-vector of 8 * 8 block 300A of loss will be positioned on this plane 500.Represent that with mathematical expression the motion-vector of 8 * 8 blocks of loss can use centre coordinate (0,0) and be expressed as:

$Z_i(x,y)={\mathrm{<figure-callout\; id="1"\; label="V"\; filenames="H2009101498028E00071.png,H2009101498028E00091.png"\; state="\{\{state\}\}">V</figure-callout>}}_1\frac{{\mathrm{\&Delta;}}_1}{\mathrm{\&Delta;}}+{\mathrm{<figure-callout\; id="2"\; label="V"\; filenames="H2009101498028E00051.png,H2009101498028E00071.png"\; state="\{\{state\}\}">V</figure-callout>}}_2\frac{{\mathrm{\&Delta;}}_2}{\mathrm{\&Delta;}}+{\mathrm{<figure-callout\; id="3"\; label="V"\; filenames="H2009101498028E00071.png"\; state="\{\{state\}\}">V</figure-callout>}}_3\frac{{\mathrm{\&Delta;}}_3}{\mathrm{\&Delta;}}-{\mathrm{<figure-callout\; id="4"\; label="V"\; filenames="H2009101498028E00071.png"\; state="\{\{state\}\}">V</figure-callout>}}_4\frac{{\mathrm{\&Delta;}}_4}{\mathrm{\&Delta;}}+{\mathrm{<figure-callout\; id="5"\; label="V"\; filenames="H2009101498028E00051.png,H2009101498028E00071.png"\; state="\{\{state\}\}">V</figure-callout>}}_5\frac{{\mathrm{\&Delta;}}_5}{\mathrm{\&Delta;}}+{\mathrm{<figure-callout\; id="6"\; label="V"\; filenames="H2009101498028E00051.png,H2009101498028E00071.png"\; state="\{\{state\}\}">V</figure-callout>}}_6\frac{{\mathrm{\&Delta;}}_6}{\mathrm{\&Delta;}}$

Wherein,

Δ＝det(M)， $M=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₁＝det(M ₁)， $M_1=\left[\begin{array}{cccccc}{x}^2& \mathrm{xy}& y^2& x& y& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₂＝det(M ₂)， $M_2=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ x^2& \mathrm{xy}& y^2& x& y& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₃＝det(M ₃)， $M_3=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ x^2& \mathrm{xy}& y^2& x& y& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₄＝det(M ₄)， $M_4=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ x^2& \mathrm{xy}& y^2& x& y& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₅＝det(M ₅)， $M_5=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ x^2& \mathrm{xy}& y^2& x& y& 1\\ {x_6}^2& x_6{y}_6& {y_6}^2& x_6& y_6& 1\end{array}\right]$

Δ ₆＝det(M ₆)， $M_6=\left[\begin{array}{cccccc}{x_1}^2& x_1{y}_1& {y_1}^2& x_1& y_1& 1\\ {x_2}^2& x_2{y}_2& {y_2}^2& x_2& y_2& 1\\ {x_3}^2& x_3{y}_3& {y_3}^2& x_3& y_3& 1\\ {x_4}^2& x_4{y}_4& {y_4}^2& x_4& y_4& 1\\ {x_5}^2& x_5{y}_5& {y_5}^2& x_5& y_5& 1\\ x^2& \mathrm{xy}& y^2& x& y& 1\end{array}\right]$

In calculating Fig. 3 behind the motion-vector of 48 * 8 block 300A, 300B, 300C and 300D; The present invention's H.264 defined variable region block size pattern of can arranging in pairs or groups; According to 4 motion-vectors that calculated, and decision need to adopt T1 pattern (16 * 16), T2 pattern (16 * 8), T3 pattern (8 * 16) or T4 (8 * 8) pattern to carry out the follow-up motion-vector program of refining.Fig. 6 shows the method flow diagram in order to decision variable region block size pattern that one embodiment of the invention provides.At first, in step S600, be example with the impaired macro zone block among Fig. 3 300, obtain 48 * 8 block 300A in the impaired macro zone block 300,300B, 300C, and the prediction motion-vector of 300D.Then, in step S610, the motion-vector of adjacent two 8 * 8 blocks of comparison level direction, and judge whether to meet the following conditions:

| MV1-MV2|≤TH _H1And | MV3-MV4|≤TH _H2

Wherein MV1, MV2, MV3 and MV4 are respectively upper left 8 * 8 blocks, upper right 8 * 8 blocks, a left side 8 * 8 blocks, and the prediction motion-vector of bottom right 8 * 8 blocks (promptly being respectively block 300A, 300D, 300B and 300C among Fig. 3) down, TH _H1And TH _H2Be the difference critical value of horizontal direction motion-vector, can adjust according to the practical application difference.In one embodiment, TH _H1And TH _H2Can all be set at 1.If satisfy above-mentioned condition, then can carry out the merging of horizontal direction, and continue to step S620, otherwise, if do not satisfy above-mentioned condition, then do not carry out the merging of horizontal direction, and program proceeds to step S630.In step S620 and S630, compare the motion-vector of adjacent two 8 * 8 blocks of vertical direction, and judge whether to meet the following conditions:

| MV1-MV3|≤TH _V1And | MV2-MV4|≤TH _V2

TH wherein _V1And TH _V2Be the difference critical value of vertical direction motion-vector, can all be set at 1 in one embodiment.If in step S620, be judged as is that then program proceeds to step S622, takes the T1 pattern; Four 8 * 8 blocks are merged formation one 16 * 16 macro zone blocks, otherwise then program proceeds to step S624; Take the T2 pattern, only carry out the merging of horizontal direction and form two 16 * 8 sub-macro zone blocks.In addition, in step S630, be that then program proceeds to step S632 if be judged as; Take the T3 pattern, carry out the merging of vertical direction and form two 8 * 16 sub-macro zone blocks, otherwise; Then program proceeds to step S634, does not carry out any merging, takes to have the T4 pattern of four 8 * 8 sub-macro zone blocks.

After pattern (T1, T2, T3 or T4 pattern) is cut apart in decision, carry out the motion-vector program of refining to each the sub-macro zone block (sub-macroblock) after cutting apart, in reference picture, to hunt out better compensation block.Refine mobile assessment (motion estimation) program of the similar coding side of program of motion-vector; With the motion-vector that each sub-macro zone block was doped is that starting point is (if sub-macro zone block comprises 8 * 8 blocks more than two; The mean value of its pairing motion-vector prediction motion-vector that is each 8 * 8 block then); Compare the error amount (like absolute error sum total (SAD)) of its peripheral boundary pixel, in reference picture, to seek better compensation block.The comparison order that the one embodiment of the invention that illustrates Fig. 7 adopts; It is starting point (being the point 0 among Fig. 7) with the motion-vector that dopes, in a search window of reference picture, adopt to order that external spiral is searched in regular turn with search window in each candidate block comparison.

In comparison process, the present invention proposes a kind of partial distortion (partial distortion) comparison method and two kinds and can refine the required time to save motion-vector in order to the method for premature termination (early termination).First kind of premature termination method is to set a critical value: DT according to the original residual value (residual) of pairing border block when encoding _a=μ * N+ γ, wherein N represents the total pixel number amount in the corresponding sides battery limit (BL) piece, and μ represents the average residual residual value of each respective pixel, and γ represents a constant coefficient.Second method is to set another critical value: DT according to the comparison result of other impaired block that has compensated in the present picture _b=EBME (u, v) * EBME _β/ EBME _α* λ+ε, wherein (u v) reaches EBME (u, the v) present pairing motion-vector of sub-macro zone block that will compare of institute and compare resulting peripheral boundary for the first time and compare error amount, EBME of representative _αAnd EBME _βRepresent the peripheral boundary comparison error amount of comparing resultant peripheral boundary comparison error amount and minimum the first time of other impaired block of before having compared in the present picture respectively; λ is constant proportionality coefficient (for example can be made as 0.6), and ε is constant coefficient (for example can be made as 0).In the process of carrying out border block comparison, when the comparison error amount is less than or equal to critical value DT _aOr DT _bThe time, the block that can mate has been found in representative, just can stop the search to this impaired sub-macro zone block, to reduce the operand of peripheral boundary comparison.In addition, partial distortion comparison method proposed by the invention is to be a unit with 16 pixels (4 * 4), and compares with the mode of one-accumulate one pixel to per unit, refines and need carry out the quantity of the right pixel of error ratio in the program to reduce motion-vector.

Fig. 8 and Fig. 9 are for T4 (8 * 8) pattern being the sketch map and the flow chart of the program of refining of example explanation one embodiment of the invention.Block 800 is upper left side 8 * 8 blocks of an impaired macro zone block, in the program of refining, will compare above it and the peripheral adjacent block 810 of left and 820 pixel error.In general, the sad value of block 800 can be expressed as:

$D=\underset{i=0}{\overset{8}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{8}{\mathrm{\&Sigma;}}}|A_{\mathrm{Top}}(x_0+i,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+j),-R(x_0+i+u,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+j+v)|+$

$\underset{i=0}{\overset{8}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{8}{\mathrm{\&Sigma;}}}|A_{\mathrm{Left}}(x_0+i,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+j),-R(x_0+i+u,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+j+v)|$

A wherein _TopAnd A _LeftRepresent the pixel in the present picture, (x y) represents the pairing pixel of reference picture, (x to R ₀, y ₀) coordinate of upper left pixel of the corresponding block of representative, (u v) represents block 800 pairing motion-vectors.

With reference to figure 8, partial distortion comparison method of the present invention is a unit with 4 * 4 blocks, and peripheral adjacent block 810 and 820 is divided into 44 * 4 blocks 812,814,816,818,822,824,826,828 respectively, and each 4 * 4 block comprises 16 pixels.The present invention is divided into 16 partial distortion (d with the total distortion value D of block 810 and 820 _p, p=1～16), each partial distortion (d wherein _p) comprise 8 pixels, and these 8 pixels lay respectively at the same position in 4 * 4 blocks 812,814,816,818,822,824,826,828.For instance, partial distortion d ₁Comprise 8 pixels representing with oblique line among Fig. 8.In one embodiment, d ₁To d ₁₆Order in each 4 * 4 block is as shown in Figure 8, and each partial distortion d _pCan represent as follows:

$d_p=\underset{i=0}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{1}{\mathrm{\&Sigma;}}}\begin{array}{c}|A_{\mathrm{Top}}(x_0+4i+s_p,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+4j+t_p),-\\ R(x_0+4i+s_p+u,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+4j+t_p+v)|\end{array}+$

$\underset{i=0}{\overset{1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{1}{\mathrm{\&Sigma;}}}\begin{array}{c}|A_{\mathrm{Left}}(x_0+4i+s_p,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+4j+t_p),-\\ R(x_0+4i+s_p+u,{\mathrm{<figure-callout\; id="0"\; label="y"\; filenames="H2009101498028E00051.png,H2009101498028E00091.png"\; state="\{\{state\}\}">y</figure-callout>}}_0+4j+t_p+v)|\end{array}$

Pairing (the s of each p value _p, t _p) value is to be listed in the table below:

p	(s p，t p)
		1	(0，0)
2	(2，2)
		3	(2，0)
4	(0，2)
		5	(1，1)
6	(3，3)
		7	(3，1)
8	(1，3)

[0063]?

9	(1，0)
		10	(3，2)
11	(0，1)
		12	(2，3)
13	(3，0)
		14	(1，2)
15	(2，1)
		16	(0，3)

In addition, definition is added to p the distortion value D that adds up of p partial distortion _pFor:

$D_p=\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{d}_i$

Below will explain in an embodiment of the present invention and carry out the program that motion-vector is refined according to above definition to an impaired block.With reference to figure 9, in step S900, judge whether the peripheral boundary comparison first time that impaired block carried out for this reason, if then proceed to step S910, calculate all partial distortion d ₁～d ₁₆And total distortion value D ₁₆=d ₁+ d ₂+ ...+d ₁₆Then, in step S912 and S914, judge D respectively ₁₆Whether smaller or equal to DT _aAnd DT _b(promptly whether satisfying the condition of premature termination) if then can proceed to step S950, finishes the searching procedure of this impaired block.If step S912 and S914 all are judged as not, then program proceeds to step S916, sets the total distortion value D of first comparison ₁₆Be D _Min, as the judgment standard of the next candidate block of comparison.Then, in step S920 to S925, carry out the peripheral boundary comparison of next candidate block, begin, calculate d respectively from p=1 _p(8 points) and D _p, if D _p＞p * D _Mim/ 16, can finish the comparison of this candidate block, and proceed to step S940.If D ₁＜D _Mim/ 16, then follow the repetition above-mentioned steps, calculate d ₂And D ₂, and compare D2 and 2 * D _Mim/ 16.The order of comparison is from p=1 to p=16, and in the comparison process in a single day D takes place _pGreater than p * D _Mim/ 16 situation, promptly in the comparison of candidate block here.When comparing, if D to p=16 ₁₆Less than D _Mim, then proceed to step S930 and S932, judge D respectively ₁₆Whether smaller or equal to DT _aAnd DT _b(promptly whether satisfying the condition of premature termination) if then can proceed to step S950, finishes the searching procedure of this impaired block.If step S930 and S932 all are judged as not, then program proceeds to step S934, resets and this time compares resulting total distortion value D ₁₆Be new D _MinValue is as the judgment standard of the next candidate block of comparison.Then; In step S940, judge whether to accomplish the comparison of last candidate block in the search window, if then get back to the peripheral boundary comparison that step S920 continues next candidate block; Then proceed to step S950 if not, finish the comparison program of refining of this impaired block.

The above is merely preferred embodiment of the present invention, is not in order to limit claim scope of the present invention; All other do not break away from the equivalence of being accomplished under the disclosed spirit and changes or modification, all should be included in the above-mentioned claim scope.

Claims (8)

1. method that is applied to the temporal error concealment in the video decode, the method includes the steps of:

A. detect an impaired macro zone block, this impaired macro zone block is made up of 48 * 8 impaired blocks;

B. obtain motion-vector around a plurality of 4 * 4 adjacent block of this impaired macro zone block;

C. to each this 8 * 8 impaired block, by in these a plurality of 4 * 4 adjacent block near the motion-vector of 64 * 4 adjacent block of this 8 * 8 impaired block, determine a prediction motion-vector of this 8 * 8 impaired block;

D. this prediction motion-vector of these 48 * 8 impaired blocks is relatively merged into 2 16 * 8 in twos with decision nonjoinder these 48 * 8 impaired blocks, with these 48 * 8 impaired blocks and is merged impaired block, these 48 * 8 impaired blocks are merged into 28 * 16 in twos merges impaired block, maybe these 48 * 8 impaired blocks all merged into 1 16 * 16 and merge impaired block;

E. to after merging this 16 * 8 merge impaired block, this 8 * 16 merges impaired block or this 16 * 16 and merges impaired block, determines one of the impaired block of each this merging to predict that motion-vector merges the mean value of the prediction motion-vector of 8 * 8 impaired blocks that impaired block comprises for this; And

F. to each this 8 * 8 impaired block, the impaired block of this 16 * 8 merging, this 8 * 16 impaired block of merging or the impaired block of this 16 * 16 merging; With this impaired block respectively or to merge pairing this prediction motion-vector of impaired block be a starting point; Carry out the pixel comparison with a plurality of reference block in the search window in the reference picture, in this search window, to seek and this impaired block corresponding reference block of being complementary of the impaired block of this merging maybe.

2. the method for temporal error concealment according to claim 1, wherein step c also comprises:

Provide an active plane of regression to be: Z _i(x, y)=α ₁x _i ²+ α ₂x _iy _i+ α ₃y _i ²+ α ₄x _i+ α ₅y _i+ α ₆

The x that brings the center point coordinate and the corresponding motion-vector of these 64 * 4 adjacent block into this active plane of regression respectively _i, y _i, and Z _iIn, in the hope of alpha ₁, α ₂, α ₃, α ₄, α ₅, and α ₆And

Determine that this prediction motion-vector is to be positioned on this active plane of regression.

3. the method for temporal error concealment according to claim 1, wherein this pixel comparison is to calculate this impaired block respectively or merge the error amount between the peripheral boundary pixel of this interior reference block of peripheral boundary pixel and this search window of impaired block.

4. the method for temporal error concealment according to claim 3, wherein this error amount is an absolute error sum total.

5. the method for temporal error concealment according to claim 3, wherein step f carries out the pixel comparison with a spirality path.

6. the method for temporal error concealment according to claim 5; Wherein carry out in the pixel comparison process with this spirality path; When the error amount of a specific reference block stops the pixel comparison during less than a critical value, and be this correspondence reference block with this specific reference block, this critical value is DT _a=μ * N+ γ, wherein N represents the total pixel number amount in this peripheral boundary pixel, and μ represents the average residual residual value of this peripheral boundary pixel, and γ represents a constant coefficient.

7. the method for temporal error concealment according to claim 5; Wherein carry out in the pixel comparison process with this spirality path; When the error amount of specific reference block stops the pixel comparison during less than critical value, and be this correspondence reference block with this specific reference block, this critical value is DT _b=EBME (u, v) * EBME _β/ EBME _α* λ+ε, wherein (u v) reaches EBME (u, v) present this impaired block that will compare of the institute pairing motion-vector of the impaired block of this merging and compare resulting error amount, EBME the first time maybe of representative respectively _αAnd EBME _βRepresentative is compared resulting error amount and minimum error values the first time of other impaired block of process comparison respectively, and λ is the constant proportionality coefficient, and ε is a constant coefficient.

8. the method for temporal error concealment according to claim 1, the pixel of wherein carrying out to each this 8 * 8 impaired block or the impaired block of each this merging is compared and is also comprised following steps:

Peripheral boundary pixel and this impaired block that calculates pairing one first reference block of this starting point be the error amount D between the peripheral boundary pixel of the impaired block of this merging maybe _MimAnd

Carry out the pixel comparison of second reference block according to a spirality path; With this impaired block maybe the peripheral boundary pixel of the impaired block of this merging be divided into 16 equal portions; Definition is a p equal portions add up error value with the error amount that the error amount of the 1st equal portions in these 16 equal portions is added in the p equal portions in regular turn; Wherein p is 1 to 16 integer, calculates this p equal portions add up error value in regular turn and judges that whether this p equal portions add up error value is greater than p * D _Mim/ 16, if can finish the pixel comparison of this second reference block.

Images