CN105681809A - Motion compensation method for double-forward prediction unit - Google Patents

Abstract

The embodiment of the invention provides a motion compensation method for a double-forward prediction unit. The method mainly comprises following steps of obtaining two initial double-forward prediction image blocks for predicting a current image block; selecting pixels on the two initial double-forward prediction image blocks; utilizing a gradient calculation formula to calculate x direction and y direction derivatives of the pixels on the forward prediction image blocks; through adoption of a training window, calculating the offset values of the pixels on the forward prediction image blocks; and adjusting the positions of the pixels on the forward prediction image blocks according to the x direction and y direction derivatives and the offset values of the pixels on the forward prediction image blocks. According to the motion compensation method for the double-forward prediction unit provided by the embodiment of the invention, the prediction effects of the prediction image block can be further improved on the basis of not increasing a code rate; the accuracies of prediction image blocks are improved; therefore, the prediction quality of the current image block is improved; and the coding efficiency of the double-forward prediction unit is improved.

Description

Motion compensation process for double; two forward prediction unit

Technical field

The present invention relates to technical field of video coding, particularly relate to a kind of motion compensation process for double; two forward prediction unit.

Background technology

Along with the extensive use of multimedia technology, and the day by day expansion of multi-medium data, the importance of video coding technique day by day highlights. Modern coding techniques adopts hybrid encoding frame, the process such as including prediction, conversion, quantization and entropy code. Predictive coding includes infra-frame prediction and inter prediction, the former utilizes the encoded image block built of laying equal stress in same two field picture that the image block currently encoded is predicted, and the latter is that the image currently encoded is predicted by the image utilizing encoded other frame laid equal stress on and build. Wherein, inter prediction encoding make use of the temporal correlation of video sequence, eliminates spatial redundancy, is very important link in current video coding framework.

The method introducing double; two forward motion compensation in latest generation video encoding standard (HEVC), under Lowdelay configures, at prediction PU (PredictionUnit, current prediction unit) time, encoder can be searched for and obtain two predicted picture blocks, and using the weighted value of the two predicted picture block as the predictive value of current PU. In the prior art, but without a kind of fine setting that predicted picture block carried out Pixel-level method to promote forecast quality further.

Summary of the invention

The embodiment provides a kind of motion compensation process for double; two forward prediction unit, to improve the accuracy of predicted picture block.

To achieve these goals, this invention takes following technical scheme.

A kind of motion compensation process for double; two forward prediction unit, including:

Obtain two forward-predicted picture blocks for predicting current image block, choose the pixel on said two forward-predicted picture block;

Gradient calculation formula is utilized to calculate x direction and the y directional derivative of pixel on described forward-predicted picture block, by training window to calculate the deviant of the pixel on described forward-predicted picture block;

According to the x direction of the pixel on described forward-predicted picture block, y directional derivative and deviant, the position of the pixel on described forward-predicted picture block is adjusted.

Further, described two the forward-predicted picture blocks obtained for predicting current image block, choose the pixel on said two forward-predicted picture block, including:

Choose the pixel p on said two forward-predicted picture block₁[i, j] and p₀[i, j], if pixel p₀Optimum prediction pixel after the adjustment of [i, j] is p'₀[i, j], pixel p'₀[i, j] is relative to pixel p₀The skew of [i, j] is (v_x, v_y), if pixel p₀Optimum prediction pixel after the adjustment of [i, j] is p'₁[i, j], pixel p'₁[i, j] is relative to pixel p₁The skew of [i, j] is (-v_x,-v_y);

According to Taylor's single order expansion formula, p'₁[i, j] and p'₀The estimated value computing formula of [i, j] is as follows:

p'₀[i,j]≈p₀[i,j]+v_x·I_x0+v_y·I_y0

p'₁[i,j]≈p₁[i,j]-v_x·I_x1-v_y·I_y1

I_x0,I_y0Represent pixel p₀The x direction of [i, j] and y directional derivative, I_x1,I_y1Represent pixel p₁The x direction of [i, j] and y directional derivative.

Further, the described x direction utilizing gradient calculation formula to calculate pixel on described forward-predicted picture block and y directional derivative, including:

I is calculated by following gradient calculation formula_x1,I_x0,I_y1,I_y0:

I_x0=(p₀[i+Δ,j]-p₀[i-Δ,j])/2

I_x1=(p₁[i+Δ,j]-p₁[i-Δ,j])/2

I_y0=(p₀[i,j+Δ]-p₀[i,j-Δ])/2

I_y1=(p₁[i,j+Δ]-p₁[i,j-Δ])/2

In formula, Δ represents image element interpolation precision set in advance, p₀[i+Δ,j]、p₀[i-Δ,j]、p₁[i+Δ,j]、p₁[i-Δ,j]、p₀[i,j+Δ]、p₀[i,j-Δ]、p₁[i,j+Δ]、p₁[i, j-Δ] is obtained by DCT interpolation filter interpolation respectively.

Further, the described deviant by training window to calculate the pixel on described forward-predicted picture block, including:

Respectively with pixel p₁[i, j] and p₀[i, j] certain contiguous range around opens a training window, utilizes method of least square to solve deviant v_x,v_y。

Wherein window training calculate computational methods particularly as follows:

$\begin{array}{c}\mathrm{\Δ}\[i,j\]={p^{\′}}_0\[i,j\]-{p^{\′}}_1\[i,j\]\\ =(p_0\[i,j\]+v_x\[i,j\]\·I_x^{\left(0\right)}\[i,j\]+v_y\[i,j\]\·I_y^{\left(0\right)}\[i,j\])\\ -(p_1\[i,j\]-v_x\[i,j\]\·I_x^{\left(1\right)}\[i,j\]-v_y\[i,j\]\·I_y^{\left(1\right)}\[i,j\])\\ \underset{\[i^{\′},j\]\∈\mathrm{\Ω}}{\mathrm{\Σ}}{\mathrm{\Δ}}^2\[i,j\]=\min .\end{array}$

By windowing, training is askedMinima obtains optimum offset m in, wherein,

$v_x\[i,j\]=\frac{{\det }_1}{\det },v_y\[i,j\]=\frac{{\det }_2}{\det },$

det₁=s₃s₅-s₂s₆,det₂=s₁s₆-s₃s₄, det=s₁s₅-s₂s₄;

$s_1=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}{(I_{x0}\[i,j\]+I_{x1}\[i,j\])}^2,$

$s_2=s_4=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(I_{x0}\[i,j\]+I_{x1}\[i,j\])(I_{y0}\[i,j\]+I_{y1}\[i,j\]),$

$s_3=-\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(P_0\[i,j\]-P_1\[i,j\])(I_{x0}\[i,j\]+I_{x1}\[i,j\]),$

$s_5=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}{(I_{y0}\[i,j\]+I_{y1}\[i,j\])}^2,$

$s_6=-\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(P_0\[i,j\]-P_1\[i,j\])(I_{y0}\[i,j\]+I_{y1}\[i,j\]).$

Further, described according to the x direction of the pixel on described forward-predicted picture block, y directional derivative and deviant, the position of the pixel on described forward-predicted picture block is adjusted, including:

In units of 2x2 pixel, offseting average using the position of adjacent four pixels as overall side-play amount, the position deviant of adjacent 4 pixels is carried out mean filter, mean filter computing formula is as follows:

v_{x_average}=(v_x1+v_x2+v_x3+v_x4)/4

v_{y_average}=(v_y1+v_y2+v_y3+v_y4)/4

(v in above-mentioned formula_X1,v_y1)、(v_X2,v_y2)、(v_x3, v_y3)、(v_x4, v_y4) the position deviant of respectively adjacent four pixels.

According to Taylor's single order expansion formula, after position deviant adjusts, two initial pixel point p₁[i, j] and p₀[i, j] is adjusted to:

p'₀[i,j]≈p₀[i,j]+v_{x_average}·I_x0+v_{y_average}·I_y0

p'₁[i,j]≈p1[i,j]-v_{x_average}·I_x1-v_{y_average}·I_y1。

Further, described method also includes:

The final predictive value of the corresponding pixel points in described current image block is adjusted to:

P_pre[i, j]=(p'₁[i,j]+p'₀[i,j])/2

RD cost after completing according to adopting the described coding for the current image block after the motion compensation process of double; two forward prediction unit, determines whether to adopt the described motion compensation process for double; two forward prediction unit.

The technical scheme provided by embodiments of the invention described above can be seen that, the motion compensation process for double; two forward prediction unit that the embodiment of the present invention proposes is by Taylor expansion and high-precision gradient calculation process, the predictive value fine setting that furthermore achieved that Pixel-level on the basis of block is predicted at original image, the prediction effect improving predicted picture block on the basis of code check further can not increased, improve the accuracy of predicted picture block, thus improving the forecast quality of current image block, improve the code efficiency of double; two forward prediction unit.

Aspect and advantage that the present invention adds will part provide in the following description, and these will become apparent from the description below, or is recognized by the practice of the present invention.

Accompanying drawing explanation

In order to be illustrated more clearly that the technical scheme of the embodiment of the present invention, below the accompanying drawing used required during embodiment is described is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the premise not paying creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings.

A kind of motion compensation process for double; two forward direction PU that Fig. 1 provides for the embodiment of the present invention realize principle schematic;

The process chart of a kind of motion compensation process for double; two forward direction PU that Fig. 2 provides for the embodiment of the present invention;

Fig. 3 is filtered schematic diagram for the position deviant of a kind of adjacent 4 pixels that the embodiment of the present invention provides.

Detailed description of the invention

Being described below in detail embodiments of the present invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has the element of same or like function from start to finish. The embodiment described below with reference to accompanying drawing is illustrative of, and is only used for explaining the present invention, and is not construed as limiting the claims.

Those skilled in the art of the present technique are appreciated that unless expressly stated, and singulative used herein " ", " one ", " described " and " being somebody's turn to do " may also comprise plural form. Should be further understood that, the wording " including " used in the description of the present invention refers to there is described feature, integer, step, operation, element and/or assembly, but it is not excluded that existence or adds other features one or more, integer, step, operation, element, assembly and/or their group. It should be understood that when we claim element to be " connected " or during " coupled " to another element, it can be directly connected or coupled to other elements, or can also there is intermediary element. Additionally, " connection " used herein or " coupling " can include wireless connections or couple. Wording "and/or" used herein includes one or more any cell listing item being associated and all combinations.

Those skilled in the art of the present technique are appreciated that unless otherwise defined, all terms used herein (include technical term and scientific terminology) and have with the those of ordinary skill in art of the present invention be commonly understood by identical meaning. Should also be understood that in such as general dictionary, those terms of definition should be understood that have the meaning consistent with the meaning in the context of prior art, and unless defined as here, will not explain by idealization or excessively formal implication.

For ease of the understanding to the embodiment of the present invention, it is further explained explanation below in conjunction with accompanying drawing for several specific embodiments, and each embodiment is not intended that the restriction to the embodiment of the present invention.

The embodiment of the present invention to solve the technical problem that the fine setting being double; two forward-predicted picture blocks carry out Pixel-level is to promote forecast quality further. By Taylor expansion and high-precision gradient calculation process, the embodiment of the present invention predicts the predictive value fine setting that furthermore achieved that Pixel-level on the basis of block at original image, do not increase the prediction effect improving prediction block on the basis of code check further, improve the code efficiency of double; two forward prediction unit.

A kind of motion compensation process for double; two forward direction PU that the embodiment of the present invention provides realize principle schematic as it is shown in figure 1, concrete handling process is as in figure 2 it is shown, include following process step:

Step S210, according to Taylor's single order expansion formula, calculate optimum prediction pixel p'₁[i, j] and p'₀The estimated value of [i, j].

For double; two forward direction PU, utilizing the search of block-based motion estimation algorithm to obtain for two the initial forward-predicted picture blocks predicting current image block, these two initial forward-predicted picture blocks are all located at before current image block on a timeline.

For the pixel p of correspondence position on two initial forward-predicted picture blocks₁[i, j] and p₀[i, j], it is assumed that the present invention needs the pixel p solved₀Optimum prediction pixel after the adjustment of [i, j] is p'₀[i, j], pixel p'₀[i, j] is positioned at pixel p₀Near [i, j], relative to initial position p₀The skew of [i, j] is (v_x, v_y). Need the pixel p solved₁Optimum prediction pixel after the adjustment of [i, j] is p'₁[i, j], pixel p'₁[i, j] is positioned at pixel p₁Near [i, j], relative to home position p₁The skew of [i, j] is (-v_x,-v_y)。

According to Taylor's single order expansion formula, p'₁[i, j] and p'₀The estimated value of [i, j] is shown below:

p'₀[i,j]≈p₀[i,j]+v_x·I_x0+v_y·I_y0

p'₁[i,j]≈p₁[i,j]-v_x·I_x1-v_y·I_y1

I_x0,I_y0Represent pixel p₀The x direction of [i, j] and y directional derivative, I_x1,I_y1Represent pixel p₁The x direction of [i, j] and y directional derivative.

Step S220, calculate derivative value I by gradient calculation formula_x1,I_x0,I_y1,I_y0。

P' is obtained in order to calculate₁[i, j] and p'₀[i, j], it is necessary to respectively obtain I_x1,I_x0,I_y1,I_y0, and v_x,v_y. Obtaining high-precision Grad to calculate, the embodiment of the present invention uses high-precision image element interpolation scheme, calculates I by following gradient calculation formula_x1,I_x0,I_y1,I_y0:

I_x0=(p₀[i+Δ,j]-p₀[i-Δ,j])/2

I_x1=(p₁[i+Δ,j]-p₁[i-Δ,j])/2

I_y0=(p₀[i,j+Δ]-p₀[i,j-Δ])/2

I_y1=(p₁[i,j+Δ]-p₁[i,j-Δ])/2

In formula, Δ represents image element interpolation precision set in advance. Wherein, p₀[i+Δ,j]、p₀[i-Δ,j]、p₁[i+Δ,j]、p₁[i-Δ,j]、p₀[i,j+Δ]、p₀[i,j-Δ]、p₁[i,j+Δ]、p₁[i, j-Δ] can be obtained by DCT (DiscreteCosineTransform, discrete cosine transform) interpolation filter interpolation respectively.

Utilizing image interpolation method to calculate and obtain 1/12 location of pixels pixel value, this programme uses 8 ladder degree interpolation filters, according to original predictive pixel present position, calculates the position Grad obtaining original predictive pixel, and interpolation filter coefficients is as follows:

Whole location of pixels :-8,19 ,-40,0,43 ,-20,12 ,-6}

1/4 location of pixels :-8,12 ,-16 ,-32,56 ,-16,12 ,-8}

1/2 location of pixels: 0,4,8 ,-52,52 ,-8 ,-4,0}

3/4 location of pixels: 8 ,-12,16 ,-56,32,16 ,-12,8}

Step S230, by train window calculate deviant v_x,v_y。

For the predictive value p' after adjusting₁[i, j] and p'₀[i, j], it is believed that difference value is the smaller the better, on this basis, respectively with pixel p₁[i, j] and p₀[i, j] certain contiguous range (such as 5x5 scope) around opens a training window, utilizes method of least square to solve deviant v_x,v_y。

Wherein window training calculate computational methods particularly as follows:

$\begin{array}{c}\mathrm{\Δ}\[i,j\]={p^{\′}}_0\[i,j\]-{p^{\′}}_1\[i,j\]\\ =(p_0\[i,j\]+v_x\[i,j\]\·I_x^{\left(0\right)}\[i,j\]+v_y\[i,j\]\·I_y^{\left(0\right)}\[i,j\])\\ -(p_1\[i,j\]-v_x\[i,j\]\·I_x^{\left(1\right)}\[i,j\]-v_y\[i,j\]\·I_y^{\left(1\right)}\[i,j\])\\ \underset{\[i^{\′},j\]\∈\mathrm{\Ω}}{\mathrm{\Σ}}{\mathrm{\Δ}}^2\[i,j\]=\min .\end{array}$

By windowing, training is askedMinima can calculate and obtain optimum deviant, and using the side-play amount of pixel total optimization in window as the optimum side-play amount of current pixel, it is possible to obtain the effect of robust more. Wherein,

$v_x\[i,j\]=\frac{{\det }_1}{\det },v_y\[i,j\]=\frac{{\det }_2}{\det },$

det₁=s₃s₅-s₂s₆,det₂=s₁s₆-s₃s₄, det=s₁s₅-s₂s₄;

$s_1=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}{(I_{x0}\[i,j\]+I_{x1}\[i,j\])}^2,$

$s_2=s_4=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(I_{x0}\[i,j\]+I_{x1}\[i,j\])(I_{y0}\[i,j\]+I_{y1}\[i,j\]),$

$s_3=-\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(P_0\[i,j\]-P_1\[i,j\])(I_{x0}\[i,j\]+I_{x1}\[i,j\]),$

$s_5=\underset{\mathrm{\Ω}}{\mathrm{\Σ}}{(I_{y0}\[i,j\]+I_{y1}\[i,j\])}^2,$

$s_6=-\underset{\mathrm{\Ω}}{\mathrm{\Σ}}(P_0\[i,j\]-P_1\[i,j\])(I_{y0}\[i,j\]+I_{y1}\[i,j\]).$

In actual applications, p'₁[i, j] is relative to p₁The skew of [i, j] position can also be expressed as (scale*v_x, scale*v_y), wherein scale represents p₁[i, j] place reference frame and present frame distance and p₀[i, j] place reference frame and present frame ratio of distances constant.

Step S240, adjacent four pixels position skew Filtering Processing

For avoiding the situation through this patent method rear section pixel prediction deleterious, we are the deviant v to adjacent 4 pixels_x,v_yBy after Filtering Processing as overall displacement trim values.

In units of 2x2 pixel, the average side-play amount as entirety is offset using the position of adjacent four pixels, the position deviant of adjacent 4 pixels is carried out mean filter, and Fig. 3 is filtered schematic diagram for the position deviant of a kind of adjacent 4 pixels that the embodiment of the present invention provides. Mean filter computational methods particularly as follows:

v_{x_average}=(v_x1+v_x2+v_x3+v_x4)/4

v_{y_average}=(v_y1+v_y2+v_y3+v_y4)/4

(v in above-mentioned formula_x1, v_y1)、(v_x2, v_y2)、(v_x3, v_y3)、(v_x4, v_y4) the position deviant of respectively adjacent four pixels.

According to Taylor's single order expansion formula, after position deviant adjusts, two initial prediction pixel p₁[i, j] and p₀[i, j] is adjusted to:

p'₀[i,j]≈p₀[i,j]+v_{x_average}·I_x0+v_{y_average}·I_y0

p'₁[i,j]≈p₁[i,j]-v_{x_average}·I_x1-v_{y_average}·I_y1

The final predictive value of the corresponding pixel points in current image block is adjusted to:

P_pre[i, j]=(p'₁[i,j]+p'₀[i,j])/2

Step S250, increase coding maker position at different levels. The predictive value of each pixel is that principle determines whether present encoding unit uses the scheme in the embodiment of the present invention with RD (rate-distortioncost, rate distortion) cost after adjusting.

The flag of flag and the PU level arranging CTU (code tree unit) level shows whether current CTU or PU uses the motion compensation process that the present invention proposes.

Method particularly as follows:

Coding side uses two schemes to carry out motion compensation for each CTU respectively, a kind of scheme is the motion compensation process that the present invention proposes, the another kind of encoding scheme using encoder original, two schemes selects optimum motion compensated schemes with RD cost for index after having encoded, if the compensation scheme RD cost that the present invention proposes is relatively low, then CTU level flag bit is set to 1, is otherwise set to 0. This flag bit is with bit stream to decoder, and decoder selects corresponding mode to be decoded end motion compensation after decoding this flag bit. Wherein,

RD-cost=R+ λ * D

Bit needed for the current CTU of R presentation code, D represents the pixel value deviation between reconstruct CTU and original CT U, and λ is constant, encoder determine.

Additionally, when CTU level flag bit is 1, for this CTU using the PU that AMVP pattern is predicted show whether this PU uses the motion compensated schemes that the present invention proposes by increasing a flag bit.

The present invention is integrated on HEVC reference software HM12.0, and for HEVC universal test sequence, when testing time 2s, the lower performance realized of Lowdelay configuration is as shown in table 1.

Experimental data shows, this programme Y-component on HM12.0 platform can obtain the performance gain of average 1.3%, sequence performance boost for having texture-rich and slight movement becomes apparent from, and such as BQsquare, under this sequence, this algorithm can obtain the performance gain of 3.9%.

Although it should be noted that the present invention is integrated in HEVC reference software HM12.0, but it can be equally applicable to other encoding and decoding platform, as H.264/AVC, AVS2 etc.

1 algorithm of table performance data under different cycle testss

In sum, the motion compensation process for double; two forward prediction unit that the embodiment of the present invention proposes is by Taylor expansion and high-precision gradient calculation process, the predictive value fine setting that furthermore achieved that Pixel-level on the basis of block is predicted at original image, the prediction effect improving predicted picture block on the basis of code check further can not increased, improve the accuracy of predicted picture block, thus improving the forecast quality of current image block, improve the code efficiency of double; two forward prediction unit.

One of ordinary skill in the art will appreciate that: accompanying drawing is the schematic diagram of an embodiment, module or flow process in accompanying drawing are not necessarily implemented necessary to the present invention.

As seen through the above description of the embodiments, those skilled in the art is it can be understood that can add the mode of required general hardware platform by software to the present invention and realize. Based on such understanding, the part that prior art is contributed by technical scheme substantially in other words can embody with the form of software product, this computer software product can be stored in storage medium, such as ROM/RAM, magnetic disc, CD etc., including some instructions with so that a computer equipment (can be personal computer, server, or the network equipment etc.) perform the method described in some part of each embodiment of the present invention or embodiment.

Each embodiment in this specification all adopts the mode gone forward one by one to describe, between each embodiment identical similar part mutually referring to, what each embodiment stressed is the difference with other embodiments. Especially for device or system embodiment, owing to it is substantially similar to embodiment of the method, so describing fairly simple, relevant part illustrates referring to the part of embodiment of the method. Apparatus and system embodiment described above is merely schematic, the wherein said unit illustrated as separating component can be or may not be physically separate, the parts shown as unit can be or may not be physical location, namely may be located at a place, or can also be distributed on multiple NE. Some or all of module therein can be selected according to the actual needs to realize the purpose of the present embodiment scheme. Those of ordinary skill in the art, when not paying creative work, are namely appreciated that and implement.

The above; being only the present invention preferably detailed description of the invention, but protection scope of the present invention is not limited thereto, any those familiar with the art is in the technical scope that the invention discloses; the change that can readily occur in or replacement, all should be encompassed within protection scope of the present invention. Therefore, protection scope of the present invention should be as the criterion with scope of the claims.

Claims (6)

1. the motion compensation process for double; two forward prediction unit, it is characterised in that including:

Obtain two forward-predicted picture blocks for predicting current image block, choose the pixel on said two forward-predicted picture block;

Gradient calculation formula is utilized to calculate x direction and the y directional derivative of pixel on described forward-predicted picture block, by training window to calculate the deviant of the pixel on described forward-predicted picture block;

According to the x direction of the pixel on described forward-predicted picture block, y directional derivative and deviant, the position of the pixel on described forward-predicted picture block is adjusted.

2. the motion compensation process for double; two forward prediction unit according to claim 1, it is characterised in that described two the forward-predicted picture blocks obtained for predicting current image block, chooses the pixel on said two forward-predicted picture block, including:

Choose the pixel p on said two forward-predicted picture block₁[i, j] and p₀[i, j], if pixel p₀Optimum prediction pixel after the adjustment of [i, j] is p'₀[i, j], pixel p'₀[i, j] is relative to pixel p₀The skew of [i, j] is (v_x, v_y), if pixel p₀Optimum prediction pixel after the adjustment of [i, j] is p'₁[i, j], pixel p'₁[i, j] is relative to pixel p₁The skew of [i, j] is (-v_x,-v_y);

According to Taylor's single order expansion formula, p'₁[i, j] and p'₀The estimated value computing formula of [i, j] is as follows:

p'₀[i,j]≈p₀[i,j]+v_x·I_x0+v_y·I_y0

p'₁[i,j]≈p₁[i,j]-v_x·I_x1-v_y·I_y1

I_x0,I_y0Represent pixel p₀The x direction of [i, j] and y directional derivative, I_x1,I_y1Represent pixel p₁The x direction of [i, j] and y directional derivative.

3. the motion compensation process for double; two forward prediction unit according to claim 2, it is characterised in that the described x direction utilizing gradient calculation formula to calculate pixel on described forward-predicted picture block and y directional derivative, including:

I is calculated by following gradient calculation formula_x1,I_x0,I_y1,I_y0:

I_x0=(p₀[i+Δ,j]-p₀[i-Δ,j])/2

I_x1=(p₁[i+Δ,j]-p₁[i-Δ,j])/2

I_y0=(p₀[i,j+Δ]-p₀[i,j-Δ])/2

I_y1=(p₁[i,j+Δ]-p₁[i,j-Δ])/2

In formula, Δ represents image element interpolation precision set in advance, p₀[i+Δ,j]、p₀[i-Δ,j]、p₁[i+Δ,j]、p₁[i-Δ,j]、p₀[i,j+Δ]、p₀[i,j-Δ]、p₁[i,j+Δ]、p₁[i, j-Δ] is obtained by DCT interpolation filter interpolation respectively.

4. the motion compensation process for double; two forward prediction unit according to claim 3, it is characterised in that the described deviant by training window to calculate the pixel on described forward-predicted picture block, including:

Respectively with pixel p₁[i, j] and p₀[i, j] certain contiguous range around opens a training window, utilizes method of least square to solve deviant v_x,v_y。

Wherein window training calculate computational methods particularly as follows:

By windowing, training is askedMinima obtains optimum offset m in, wherein,

det₁=s₃s₅-s₂s₆,det₂=s₁s₆-s₃s₄, det=s₁s₅-s₂s₄;

。

5. the motion compensation process for double; two forward prediction unit according to claim 4, it is characterized in that, described according to the x direction of the pixel on described forward-predicted picture block, y directional derivative and deviant, the position of the pixel on described forward-predicted picture block is adjusted, including:

In units of 2x2 pixel, offseting average using the position of adjacent four pixels as overall side-play amount, the position deviant of adjacent 4 pixels is carried out mean filter, mean filter computing formula is as follows:

v_{x_average}=(v_x1+v_x2+v_x3+v_x4)/4

v_{y_average}=(v_y1+v_y2+v_y3+v_y4)/4

(v in above-mentioned formula_x1, v_y1)、(v_x2, v_y2)、(v_x3, v_y3)、(v_x4, v_y4) the position deviant of respectively adjacent four pixels.

According to Taylor's single order expansion formula, after position deviant adjusts, two initial pixel point p₁[i, j] and p₀[i, j] is adjusted to:

p'₀[i,j]≈p₀[i,j]+v_{x_average}·I_x0+v_{y_average}·I_y0

p'₁[i,j]≈p₁[i,j]-v_{x_average}·I_x1-v_{y_average}·I_y1。

6. the motion compensation process for double; two forward prediction unit according to claim 5, it is characterised in that described method also includes:

The final predictive value of the corresponding pixel points in described current image block is adjusted to:

P_pre[i, j]=(p'₁[i,j]+p'₀[i,j])/2

RD cost after completing according to adopting the described coding for the current image block after the motion compensation process of double; two forward prediction unit, determines whether to adopt the described motion compensation process for double; two forward prediction unit.