CN104935938A - Inter-frame prediction method in hybrid video coding standard - Google Patents
Inter-frame prediction method in hybrid video coding standard Download PDFInfo
- Publication number
- CN104935938A CN104935938A CN201510415613.6A CN201510415613A CN104935938A CN 104935938 A CN104935938 A CN 104935938A CN 201510415613 A CN201510415613 A CN 201510415613A CN 104935938 A CN104935938 A CN 104935938A
- Authority
- CN
- China
- Prior art keywords
- contiguous
- block
- motion vector
- movable information
- division unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
An inter-frame prediction method in a hybrid video coding standard belongs to the field of video coding. In order to effectively process deformation movement existing in a video sequence, the present invention puts forward an inter-frame prediction method in a hybrid video coding standard for further improving coding performance of a video. The inter-frame prediction method comprises the steps of: obtaining motion information of a plurality of adjacent coded blocks around a current coding block; obtaining a reference index of each dividing unit in the current coding block according to obtained reference indexes of the adjacent coded blocked; and processing motion vectors of the adjacent coded blocks according to the obtained reference indexes of the adjacent coded blocks and the obtained reference index of each dividing unit in the current coding block, so as to obtain a motion vector of each dividing unit in the current coding block. According to the inter-frame prediction method of the present invention, motion information of the current coding block is predicted through the motion information of the adjacent coded blocks of the current coding block, so that deformation movement existing in the video sequence can be effectively described, and coding efficiency can be further improved.
Description
Technical field
The present invention relates to inter-frame prediction method in a kind of hybrid video coding standard.
Background technology
Along with the raising that people require video display quality, the new video application forms such as high definition and ultra high-definition video arise at the historic moment.Under this high-resolution high-quality video appreciation application more and more widely situation, how augmented video compression efficiency becomes most important.
Image and video, in digitized process, create a large amount of data redundancies, and this makes video compression technology become possibility.Generally speaking, redundancy type at least comprises spatial redundancy, time redundancy, comentropy redundancy.For the elimination of time redundancy, the general method adopted based on prediction, i.e. inter prediction encoding.Its basic thought is the block finding current block to mate most from encoded consecutive frame, only residual sum and motion vector need be passed to decoding end, significantly reduce the temporal redundancy in Video coding.What adopt due to existing video encoding standard is block-based motion estimation based on translational motion model, all pixels in its hypothesis block all experience identical motion, even current state-of-the-art HEVC coding standard, also cannot describe the scene of non-translational motion accurately.
In order to overcome the shortcoming of block matching algorithm, Ouseb Lee, Yao Wang, " Motion-Compensated PredictionUsing Nodal-Based Deformable Block Matching; " Journal of Visual Communication and ImageRepresentation, vol.6, no.1, pp.26-34, Mar.1995 proposes to use the distorted pattern based on node, the Controlling vertex of the given number of its hypothesis block can move freely, and in block, the MV of any point can by the MV interpolation of Controlling vertex out.The method is more responsive to initial value, and needs a large amount of calculating.
Yao Wang, J.Ostermann, " Comparison of Block-Based and Mesh-Based Motion EstimationAlgorithms; " IEEE International Symposium on Circuits and Systems, vol.2, pp.1157-1160, Jun.1997 once pointed out that the model based on grid in time there is non-translational motion (such as end rotation) was better than block-based forecast result of model.But when a block comprises multiple object towards different directions motion time, the prediction effect of these two models is all bad.
In order to Simulation of Complex motion better, Roman C.Kordasiewicz, Michael D.Gallant, and ShahramShirani, " Affine Motion Prediction Based on Translational Motion Vectors; " IEEE Trans.CircuitsSyst.Video Technol., vol.17, no.10, pp.1388-1394, Oct.2007 proposes an affine motion field prediction based on translational motion vector, after it is in mode decision and estimation, as a post-processing step.
In order to describe compound movement better, A.Glantz, M.Tok, A.Krutz, and T.Sikora, " A block adaptiveskip mode for inter prediction based on parametric motion models, " IEEE InternationalConference on Image Processing, pp.1201-1204, Sept.2011 propose a parametrization skip pattern based on high order parameters motion model.
With the method proposed in the past unlike, our scheme does not need encoded motion parameter when process amoeboid movement but utilizes the movable information of adjacent coded block to obtain the movable information of current deformation block, and can select an optimum forecast model in multiple forecast model.
Summary of the invention
The object of the invention is to effectively process the amoeboid movement existed in video sequence, and propose inter-frame prediction method in a kind of hybrid video coding standard, to promote the coding efficiency of video further.
The present invention solves the problems of the technologies described above the technical scheme taked to be:
Inter-frame prediction method in a kind of hybrid video coding standard, described Forecasting Methodology is for describing the amoeboid movement existed in video sequence, and described Forecasting Methodology is used in merge pattern, skip pattern or inter pattern;
The implementation procedure of described Forecasting Methodology is:
Step one: the movable information obtaining several adjacent coded blocks of surrounding of present encoding block, present encoding block is of a size of W*H, and W is the wide of present encoding block, and H is the height of present encoding block; Described movable information comprises reference key and motion vector; Around several adjacent coded blocks are called contiguous encoding block;
Step 2: the reference key obtaining each division unit in present encoding block according to the reference key of the contiguous encoding block of step one acquisition;
Step 3: process according to the motion vector of reference key to contiguous encoding block of each division unit in the present encoding block that reference key and the step 2 of the contiguous encoding block of step one acquisition obtain, obtain the motion vector of each division unit in present encoding block.
In step one, described contiguous encoding block is the contiguous block of the contiguous block of four corner location of present encoding block or four corner location of present encoding block and center position; Described contiguous encoding block is arranged in the spatial domain contiguous block of present frame or is arranged in the time domain contiguous block of time domain reference frame;
In step 2 and step 3, the size of division unit is w*h, wherein: 1≤w≤W, 1≤h≤H;
The principle obtaining the reference key of each division unit in present encoding block described in step 2 is:
The maximum reference key of the access times of contiguous encoding block obtained in selecting step one as object reference index,
Or for each contiguous encoding block specifies weights, added up by the weights of the contiguous encoding block with same reference index, the reference key of the maximum weight of the contiguous encoding block obtained in selecting step one is as object reference index;
Described in step 3, the motion vector of contiguous encoding block is processed, its process is: before the motion vector interpolation of use contiguous block goes out the motion vector of target division unit in current block, carry out preliminary treatment to the motion vector of contiguous block, if namely the reference key of contiguous block follows obtained object reference index different, just the motion vector of contiguous block zoomed to the motion vector corresponding to object reference index according to time domain distance.
In step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is:
Described motion vector is calculated by bilinear interpolation model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block; To choosing the motion vector of contiguous block to carry out preliminary treatment, then just can according to formula
calculate the motion vector of current block target division unit, in formula, B
mrepresent present encoding block, U
xrepresent certain division unit in current block, d
m(U
x) represent the motion vector MV of target division unit in current block, d
m(MP
k) represent B
mthe MV of a kth contiguous block, interpolation kernel φ
m,kx () depends on B
ma kth contiguous block to U
xcontribution, can be determined by the position of target division unit relative to each control MP.
In step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by the affine model of six parameters, its computational process is: choose three contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block; To choosing the motion vector of contiguous block to carry out preliminary treatment, utilize the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate six affine transformation coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6be affine transformation coefficient, at least need the motion vector of three contiguous blocks could determine described affine transformation coefficient.
In step 3, the implementation procedure obtaining the motion vector of each division unit in present encoding block is: calculated by the affine model of four parameters, its computational process is: choose two contiguous blocks in the some contiguous blocks from step one, and their movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block; To choosing the motion vector of contiguous block to carry out preliminary treatment, utilize the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate four affine transformation coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
0, a
1, a
2, a
3be affine transformation coefficient, at least need the motion vector of two contiguous blocks could determine described affine transformation coefficient.
In step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by perspective projection model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block; To choosing the motion vector of contiguous block to carry out preliminary treatment, utilize the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate eight perspective projection coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6, a
7, a
8be perspective projection coefficient, at least need the motion vector of four contiguous blocks could determine described perspective projection coefficient.
In step 3, being embodied as of motion vector of each division unit in described acquisition present encoding block: for current block selects an optimum forecast model in described bilinear interpolation model, perspective projection model, six parameter affine models, four-parameter equation.
The selection course of optimum forecast model is: choose four contiguous blocks in the some contiguous blocks obtained from step one, the parameter of bilinear interpolation model is calculated according to its relative position, four contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of perspective projection model is calculated according to its relative position and movable information, three contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of six parameter affine models is calculated according to its relative position and movable information, two contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of four-parameter equation is calculated according to its relative position and movable information, then in the set of all effective forecast model formations, select an optimum forecast model or form in the random subset of set at all effective forecast models to select an optimum forecast model, the motion vector of each division unit in present encoding block can be obtained after determining forecast model,
The motion vector of each division unit in described acquisition current block, when selecting optimum forecast model, can be candidate prediction model sequence in candidate prediction model set according to the estimated performance of forecast model or utilization rate.
The decision criteria of selection optimum prediction model is: adopt least mean-square error, minimal error quadratic sum, minimum Hadamard error or rate-distortion optimization criterion.
First being divided into N part to acquisition present encoding block, is the corresponding application of every part a kind of forecast model wherein, wherein N >=1.
The invention has the beneficial effects as follows:
Forecasting Methodology of the present invention can describe the amoeboid movement existed in video sequence effectively.
The present invention is the movable information by utilizing bilinear interpolation model, perspective projection model, six parameter affine models or four-parameter equation to obtain each division unit of present encoding block by the movable information of the adjacent coded block of current block, thus inter prediction performance is got a promotion.
The present invention can process the amoeboid movement existed in video sequence effectively, compared with block matching algorithm existing in current encoder standard, our scheme has obvious performance gain to the sequence comprising amoeboid movement, generally speaking, performance loss even also some gain be there is no to the sequence not comprising amoeboid movement.Described inter-frame prediction method may be used in merge pattern, skip pattern or inter pattern.
In inter prediction, Motion estimation and compensation can reduce the temporal redundancy in Video coding effectively, improves video coding performance.What adopt due to existing video encoding standard is block-based motion estimation based on translational motion model, even current state-of-the-art HEVC coding standard, also cannot describe accurately the scene of non-translational motion.For this reason, we propose an inter-frame prediction method based on bilinear interpolation model, affine model and perspective model, the movable information of the movable information of several adjacent coded blocks around present encoding block to current block is utilized to predict, the amoeboid movement existed in video sequence (is such as rotated, convergent-divergent etc.) effectively described, code efficiency is further enhanced.
Accompanying drawing explanation
Fig. 1 is the position candidate of current block four angle point control MP in embodiment of the present invention method, and wherein (a) represents the position candidate of upper left, upper right, lower left corner point control MP, is the spatial domain contiguous block being positioned at present frame; B () represents the position candidate of lower right corner point control MP, be the time domain contiguous block being arranged in time domain reference frame.
Fig. 2 is the bilinear interpolation model schematic based on pixel in embodiment of the present invention method.
Embodiment
Embodiment one: in the hybrid video coding standard described in present embodiment, inter-frame prediction method is for describing the amoeboid movement existed in video sequence, described Forecasting Methodology is used in merge pattern, skip pattern or inter pattern; The implementation procedure of described Forecasting Methodology is:
Step one: the movable information obtaining several adjacent coded blocks of surrounding of present encoding block, present encoding block is of a size of W*H, and W is the wide of present encoding block, and H is the height of present encoding block; Described movable information comprises reference key and motion vector; Around several adjacent coded blocks are called contiguous encoding block;
Step 2: the reference key obtaining each division unit in present encoding block according to the reference key of the contiguous encoding block of step one acquisition;
Step 3: process according to the motion vector of reference key to contiguous encoding block of each division unit in the present encoding block that reference key and the step 2 of the contiguous encoding block of step one acquisition obtain, obtain the motion vector of each division unit in present encoding block.
Embodiment two: inter-frame prediction method in the hybrid video coding standard described in present embodiment, is characterized in that:
In step one, described contiguous encoding block is the contiguous block of the contiguous block of four corner location of present encoding block or four corner location of present encoding block and center position; Described contiguous encoding block is arranged in the spatial domain contiguous block of present frame or is arranged in the time domain contiguous block of time domain reference frame;
Also support to select the contiguous block of more more number to obtain the movable information of present encoding block, except four corner location and center position, the contiguous block of other positions is supported too;
Such as, the block of what four described adjacent coded blocks of angle point were chosen is present encoding block upper left, upper right, lower-left, the contiguous 4x4 size in the lower right corner, is designated as control MP (minimum partition); Wherein, control MP in the lower right corner is arranged in time domain reference frame;
In step 2 and step 3, the size of division unit is w*h, and wherein: 1≤w≤W, 1≤h≤H, the size of division unit can be unified, also can consider that the texture of current block or other features are set to different;
The principle obtaining the reference key of each division unit in present encoding block described in step 2 is:
The maximum reference key of the access times of contiguous encoding block obtained in selecting step one as object reference index (reference key of each division unit),
Or for each contiguous encoding block specifies weights, added up by the weights of the contiguous encoding block with same reference index, the reference key of the maximum weight of the contiguous encoding block obtained in selecting step one is as object reference index;
Described in step 3, the motion vector of contiguous encoding block is processed, its process is: before the motion vector interpolation of use contiguous block goes out the motion vector of target division unit in current block, carry out preliminary treatment to the motion vector of contiguous block, if namely the object reference index that obtains with present embodiment of the reference key of contiguous block is different, just the motion vector of contiguous block zoomed to the motion vector corresponding to object reference index according to time domain distance.Other composition and annexation identical with embodiment one.
Such as: in the above-described embodiment, the reference key of each division unit in described acquisition current block, certain rule (such as, access times are maximum) can be utilized to choose object reference index according to the access times of the reference key of the control MP obtained in step one; Weights can be specified for each control MP, the weights of the control MP with same reference index are added up, weights according to the reference key of the control MP obtained in step one utilize certain rule (such as, maximum weight) to choose object reference index; For bilinear interpolation model, the weights of control MP can be bilinear interpolation models is the weights that each control MP calculates.
Such as: in the above-described embodiment, in described acquisition present encoding block, the motion vector of each division unit carried out preliminary treatment to the motion vector of control MP before the motion vector interpolation of use control MP goes out the motion vector of target division unit in current block, if namely the reference key of control MP is different with the reference key of the target division unit obtained in step 2, just the motion vector of control MP zoomed to the motion vector corresponding to object reference index according to time domain distance.
Embodiment three: inter-frame prediction method in the hybrid video coding standard described in present embodiment, in step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is:
Described motion vector is calculated by bilinear interpolation model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in embodiment two; According to embodiment two, mode carries out preliminary treatment to choosing the motion vector of contiguous block, then just can according to formula
calculate the motion vector of current block target division unit, in formula, B
mrepresent present encoding block, U
xrepresent certain division unit in current block, d
m(U
x) represent the motion vector MV of target division unit in current block, d
m(MP
k) represent B
mthe MV of a kth contiguous block, interpolation kernel φ
m,kx () depends on B
ma kth contiguous block to U
xcontribution, can be determined by the position of target division unit relative to each control MP;
With w=1, h=1 for example, the value of interpolation kernel is as follows:
φ
m,1(x)＝(W+4-dx)·(H+4-dy)/((W+4)·(H+4))
φ
m,2(x)＝dx·(H+4-dy)/((W+4)·(H+4))
φ
m,3(x)＝(W+4-dx)·dy/((W+4)·(H+4))
φ
m,4(x)＝dx·dy/((W+4)·(H+4))
In formula, W represents the wide of current block, H represents the height of current block, and dx represents in current block the center pixel pixel distance in the horizontal direction of object pixel and the left side control MP wanting interpolation, and dy represents the center pixel pixel distance in the vertical direction of object pixel and the top control MP wanting interpolation.Other composition and annexation identical with embodiment two.
Embodiment four: inter-frame prediction method in the hybrid video coding standard described in present embodiment, in step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by the affine model of six parameters, its computational process is: choose three contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in embodiment two; According to embodiment two, mode carries out preliminary treatment to choosing the motion vector of contiguous block, utilizes the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate six affine transformation coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6be affine transformation coefficient, at least need the motion vector of three contiguous blocks could determine described affine transformation coefficient.Other composition and annexation identical with embodiment two.
Embodiment five: inter-frame prediction method in the hybrid video coding standard described in present embodiment, in step 3, the implementation procedure obtaining the motion vector of each division unit in present encoding block is: calculated by the affine model of four parameters, its computational process is: choose two contiguous blocks in the some contiguous blocks from step one, and their movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in embodiment two; To choosing the motion vector of contiguous block to carry out preliminary treatment according to embodiment two, utilize the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate four affine coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
0, a
1, a
2, a
3be affine transformation coefficient, at least need the motion vector of two contiguous blocks could determine described affine transformation coefficient.Other composition and annexation identical with embodiment two.
Embodiment six: inter-frame prediction method in the hybrid video coding standard described in present embodiment, in step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by perspective projection model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in embodiment two; Mode described in embodiment two carries out preliminary treatment to choosing the motion vector of contiguous block, utilizes the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate eight perspective projection coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6, a
7, a
8be perspective projection coefficient, at least need the motion vector of four contiguous blocks could determine described perspective projection coefficient.Other composition and annexation identical with embodiment two.
Embodiment seven: inter-frame prediction method in the hybrid video coding standard described in present embodiment, in step 3, being embodied as of motion vector of each division unit in described acquisition present encoding block: for current block selects an optimum forecast model in described bilinear interpolation model, perspective projection model, six parameter affine models, four-parameter equation.Other composition and annexation identical with embodiment one or two.
Embodiment eight: inter-frame prediction method in the hybrid video coding standard described in present embodiment, the selection course of optimum forecast model is: choose four contiguous blocks in the some contiguous blocks obtained from step one, the parameter of bilinear interpolation model is calculated according to its relative position, four contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of perspective projection model is calculated according to its relative position and movable information, three contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of six parameter affine models is calculated according to its relative position and movable information, two contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of four-parameter equation is calculated according to its relative position and movable information, then in the set of all effective forecast model formations, select an optimum forecast model or form in the random subset of set at all effective forecast models to select an optimum forecast model, the motion vector of each division unit in present encoding block can be obtained after determining forecast model,
The motion vector of each division unit in described acquisition current block, when selecting optimum forecast model, can be candidate prediction model sequence in candidate prediction model set according to the estimated performance of forecast model or utilization rate.
Other composition and annexation identical with embodiment seven.
Embodiment eight is illustrated:
The motion vector of each division unit in described acquisition current block, be the contiguous block of four corner location for the some contiguous blocks of surrounding in step one, six parameter affine models can have six kinds of compound modes: 1) use upper left, upper right, six parameter affine models of lower-left angle point contiguous block movable information, 2) use upper left, upper right, six parameter affine models of bottom right angle point contiguous block movable information, 3) use upper right, lower-left, six parameter affine models of bottom right angle point contiguous block movable information, 4) use upper left, lower-left, six parameter affine models of bottom right angle point contiguous block movable information, 5) present encoding block is divided into two triangles, adopt and use upper left, upper right, six parameter affine model and upper lefts of bottom right angle point contiguous block movable information, lower-left, six parameter affine models of bottom right angle point contiguous block movable information, 6) present encoding block is divided into two triangles, adopt and use upper left, upper right, six parameter affine model and upper rights of lower-left angle point contiguous block movable information, lower-left, six parameter affine models of bottom right angle point contiguous block movable information, four-parameter equation has eight kinds of compound modes: 1) use upper right, the four-parameter equation of lower-left angle point contiguous block movable information, 2) upper left is used, the four-parameter equation of lower-left angle point contiguous block movable information, 3) upper left is used, the four-parameter equation of upper right angle point contiguous block movable information, 4) upper right is used, the four-parameter equation of bottom right angle point contiguous block movable information, 5) lower-left is used, the four-parameter equation of bottom right angle point contiguous block movable information, 6) upper left is used, the four-parameter equation of bottom right angle point contiguous block movable information, 7) present encoding block is divided into two rectangles, adopt and use upper left, the four-parameter equation of upper right angle point contiguous block movable information and lower-left, the four-parameter equation of bottom right angle point contiguous block movable information, 8) present encoding block is divided into two rectangles, adopt and use upper left, the four-parameter equation of lower-left angle point contiguous block movable information and upper right, the four-parameter equation of bottom right angle point contiguous block movable information.That is the forecast model that optimum can be selected in bilinear interpolation model, perspective projection model and above-mentioned 14 kinds of forecast models.Also the forecast model that optimum can be selected in the random subset of these 16 kinds of forecast models.The optimal prediction modes chosen carries out encoding to ensure that decoding end can normal decoder.
Embodiment nine: inter-frame prediction method in the hybrid video coding standard described in present embodiment, it is characterized in that: the decision criteria of selection optimum prediction model is: adopt least mean-square error (Mean Squared Error, MSE), minimal error square (The sum of squares due to error, SSE), minimum Hadamard error or rate-distortion optimization criterion.Other composition and annexation identical with embodiment seven or eight.
Embodiment ten: inter-frame prediction method in hybrid video coding standard described in present embodiment, is characterized in that: be first divided into N part to acquisition present encoding block is the corresponding application of every part a kind of forecast model wherein, wherein N >=1.
Such as, current block can be divided into two parts, wherein part application bilinear interpolation model, another part is forecast model optimum for current block-size selection one in bilinear interpolation model, perspective projection model, six parameter affine models, four-parameter equation; Current block can be divided into four parts, wherein some part uses six parameter affine models, and some part uses four-parameter equation.
Other composition and annexation and embodiment three, four, five, six or seven identical.
Embodiment one:
Provide the specific implementation step of inter-frame prediction method in hybrid video coding standard:
Step one: the movable information obtaining present encoding block (be of a size of W*H, W is the wide of present encoding block, and H is the height of present encoding block) four adjacent coded blocks of angle point;
Step 2: the reference key obtaining each division unit in present encoding block according to the movable information of the contiguous encoding block of step one acquisition;
Step 3: the motion vector obtaining each division unit in present encoding block according to the movable information of the contiguous encoding block of step one acquisition.
The block of what four described adjacent coded blocks of angle point were chosen is present encoding block upper left, upper right, lower-left, the contiguous 4x4 size in the lower right corner, is designated as control MP (minimum partition).Wherein, control MP in the lower right corner is arranged in time domain reference frame.As shown in Figure 1, the position candidate of upper left corner control MP is B
2, B
3, A
2; The position candidate in the upper right corner is B
0, B
1; The position candidate in the lower left corner is A
0, A
1; The position candidate in the lower right corner is T
rb.We select a control MP for each angle point, and are that each reference frame lists obtains separately control MP according to following priority:
1) for upper left corner control MP, first we check position B
0if not exist or its predictive mode is frame mode, we will check position B
1; Otherwise B
0be used as control MP.
2) for the lower left corner, inspection priority is A
0, A
1.
3) for the upper left corner, inspection priority is B
2, A
2, B
3.
In this embodiment, employing is bilinear interpolation model, six parameter affine models, four-parameter equation.
Based on pixel bilinear interpolation model schematic as shown in Figure 2, i.e. the 1*1 of the size of division unit, the bilinear interpolation core corresponding to Fig. 2 is as follows:
φ
m,1(x)＝(W+4-dx)·(H+4-dy)/((W+4)·(H+4))
φ
m,2(x)＝dx·(H+4-dy)/((W+4)·(H+4))
φ
m,3(x)＝(W+4-dx)·dy/((W+4)·(H+4))
φ
m,4(x)＝dx·dy/((W+4)·(H+4))
In formula, W represents the wide of current block, and H represents the height of current block, and dx represents in current block the object pixel and left side control MP (MP of wanting interpolation
1) center pixel pixel distance in the horizontal direction, dy represents the object pixel and top control MP (MP of wanting interpolation
2) center pixel pixel distance in the vertical direction.
The reference key of each division unit in described acquisition current block, for bilinear interpolation model, bilinear interpolation model is utilized to calculate the weights of each control MP according to target division unit relative to the position of each control MP, the weights of the control MP with same reference index are added up, choose that reference key of maximum weight, if the reference key with maximum weights has multiple, that reference key just selecting index value minimum; For six parameter affine models, four-parameter equation, from the reference key that control MP uses, choose the reference key of the maximum reference key of access times as target division unit.
The motion vector of each division unit in described acquisition current block, before the motion vector interpolation of use control MP goes out the motion vector of target division unit in current block, preliminary treatment is carried out to the motion vector of control MP, if namely the reference key of control MP to catch up with the reference key stating obtained target division unit different, just the motion vector of control MP zoomed to the motion vector corresponding to object reference index according to time domain distance.
The motion vector of each division unit in described acquisition current block, at bilinear interpolation model, six parameter affine models, for current block selects an optimum forecast model in four-parameter equation, wherein, the six parameter affine models used comprise: 1) use upper left, upper right, six parameter affine models of lower left corner point control MP movable information, 2) upper left is used, upper right, six parameter affine models of lower right corner point control MP movable information, 3) upper right is used, lower-left, six parameter affine models of lower right corner point control MP movable information, 4) upper left is used, lower-left, six parameter affine models of lower right corner point control MP movable information, the four-parameter equation used comprises: 1) use upper right, the four-parameter equation of lower left corner point control MP movable information, 2) upper left is used, the four-parameter equation of lower left corner point control MP movable information, 3) upper left is used, the four-parameter equation of upper right corner point control MP movable information, 4) upper right is used, the four-parameter equation of lower right corner point control MP movable information, 5) lower-left is used, the four-parameter equation of lower right corner point control MP movable information, 6) upper left is used, the four-parameter equation of lower right corner point control MP movable information, that is 11 kinds of optional forecast models are had.The decision criteria selecting optimum prediction model is minimum Hadamard error.Described inter-frame prediction method is used in merge pattern and skip pattern.And described inter-frame prediction method is denoted as enhancement mode, specifically describe as follows:
Step one: first build an enhancement mode candidate list, the size of this list is 8, this list is formed by following order of candidates:
1) four-parameter equation of upper right, lower left corner point control MP movable information is used
2) four-parameter equation of upper left, lower left corner point control MP movable information is used
3) six parameter affine models of upper left, upper right, lower left corner point control MP movable information are used
4) four-parameter equation of upper left, upper right corner point control MP movable information is used
5) four-parameter equation of upper right, lower right corner point control MP movable information is used
6) four-parameter equation of lower-left, lower right corner point control MP movable information is used
7) four-parameter equation of upper left, lower right corner point control MP movable information is used
8) the bilinear interpolation model of upper left, upper right, lower-left, lower right corner point control MP movable information is used
9) six parameter affine models of upper left, upper right, lower right corner point control MP movable information are used
10) six parameter affine models of upper right, lower-left, lower right corner point control MP movable information are used
11) six parameter affine models of upper left, lower-left, lower right corner point control MP movable information are used
When candidate's number available in list more than 8 time, candidate below can give up.
Step 2: choose an optimum forecast model according to minimum Hadamard error criterion
Step 3: forecast model and other patterns (common merge pattern, skip pattern, inter pattern and intra pattern) of the optimum that step 2 is chosen carry out rate-distortion optimization selection.
Only the pattern of proposal is used in merge pattern and skip pattern in the present embodiment, so without the need to coded motion information as common inter pattern.But need the optimal prediction modes chosen to be encoded, to ensure that decoding end can normal decoder.In order to reach compromise between complexity and code efficiency, the enhancement mode of proposal is only used for the coding unit being greater than or equal to 16x16 by us.
In the present embodiment, division unit is pixel.
Embodiment two:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of the present embodiment, only uses the bilinear interpolation model of upper left, upper right, lower-left, lower right corner point control MP movable information.
Embodiment three:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the perspective projection model of upper left, upper right, lower-left, lower right corner point control MP movable information.
Embodiment four:
With tool embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, present encoding block is divided into two triangles, adopts the six parameter affine models using upper left, upper right, six parameter affine models of lower right corner point control MP movable information and upper left, lower-left, lower right corner point control MP movable information.
Embodiment five:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, present encoding block is divided into two triangles, adopts the six parameter affine models using upper left, upper right, six parameter affine models of lower left corner point control MP movable information and upper right, lower-left, lower right corner point control MP movable information.
Embodiment six:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, present encoding block is divided into two rectangles, adopts the four-parameter equation using upper left, the four-parameter equation of upper right corner point control MP movable information and lower-left, lower right corner point control MP movable information.
Embodiment seven:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, present encoding block is divided into two rectangles, adopts the four-parameter equation using upper left, the four-parameter equation of lower left corner point control MP movable information and upper right, lower right corner point control MP movable information.
Embodiment eight:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses six parameter affine models of upper left, upper right, lower left corner point control MP movable information.
Embodiment nine:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses six parameter affine models of upper left, upper right, lower right corner point control MP movable information.
Embodiment ten:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses six parameter affine models of upper right, lower-left, lower right corner point control MP movable information.
Embodiment 11:
With tool embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses six parameter affine models of upper left, lower-left, lower right corner point control MP movable information.
Embodiment 12:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of upper right, lower left corner point control MP movable information.
Embodiment 13:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of upper left, lower left corner point control MP movable information.
Embodiment 14:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of upper left, upper right corner point control MP movable information.
Embodiment 15:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of upper right, lower right corner point control MP movable information.
Embodiment 16:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of lower-left, lower right corner point control MP movable information.
Embodiment 17:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, only uses the four-parameter equation of upper left, lower right corner point control MP movable information.
Embodiment 18:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, the size of complex patterns candidate list is 11.
Embodiment 19:
With embodiment one unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, division unit is the block of 4*4.
Embodiment 20:
With embodiment 19 unlike, inter-frame prediction method in the hybrid video coding standard of present embodiment, is only used for the predicting unit partition mode of 2N*2N by the enhancement mode of proposal.
The division unit of above-mentioned specific embodiment also can be the block of w*h size, and wherein, 1≤w≤W, 1≤h≤H, the size of division unit can be unified, also can consider that the texture of current block or other features are set to different.The decision criteria selecting optimal prediction modes also can be least mean-square error (Mean Squared Error, MSE), minimal error quadratic sum (The sum of squares due to error, SSE), rate-distortion optimization criterion, or other weigh the criterion of error size.Above-mentioned embodiment is also applicable to less coding unit.Except merge pattern and skip pattern, above-mentioned embodiment is equally applicable to inter pattern.Contiguous block position in above-mentioned embodiment is not limited to four corner location.
Above content is in conjunction with concrete preferred implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.For this person of an ordinary skill in the technical field, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to scope of patent protection of the present invention.
Embodiment one is in the upper realization of VC-0.2 (with the addition of the test model of some technology at the test model HM12.0 of HEVC), and survey condition test according to VC266 is logical, VC266 is logical surveys conditioned reference VC266Study Group, " Test conditionand evaluation methodology ", VC-02-N005, VC266 2th Meeting:Suzhou, Mar.2015. because our scheme is for effectively describing the amoeboid movement existed in video sequence, therefore we are also tested for the sequence that some comprise amoeboid movement, table 1 is described these Deformation Series.
The experimental result of embodiment one is as shown in table 2, as shown in Table 2, compared with VC-0.2, at Low delay BMain_HighBitrate (LB-HR), Low delay B Main_LowBitrate (LB-LR), Random AccessMain_HighBitrate (RA-HR), under Random Access Main_LowBitrate (RA-LR) configuration condition, Deformation Series on average has 8.2%, 9.9%, 6.9%, the BD bit rate of 8.2% is saved, the category-B sequence average defined in logical survey condition has 0.9%, 1.3%, 1.4%, the BD bit rate of 1.8% is saved, TMZ class sequence average has 0.4%, 0.4%, 0.3%, 0.3% BD bit rate save.BD bit rate represents that the code check of two kinds of methods under same objective quality saves situation, with reference to G.
" Calculation of average PSNR differences between RD-Curves, " ITU-T SG16 Q.6 Document, VCEG-M33, Austin, US, April 2001.
Table 1. has the sequence of amoeboid movement
| Sequence | Resolution | Frame number | Frame per second | The motion feature of sequence |
| CactusPart | 1920x1080 | 100 | 50fps | Comprise the rotation of local |
| BlueSkyPart | 1920x1080 | 100 | 25fps | Camera rotates, a lot of details |
| StationPart | 1920x1080 | 100 | 25fps | Shorten focal length significantly, a lot of details |
| TractorPart | 1920x1080 | 100 | 25fps | From local detail to overall scenario |
| ShieldsPart | 1920x1080 | 100 | 50fps | Elongate focal length significantly, a lot of details |
| SpincalendarPart | 1280x720 | 100 | 50fps | Integral-rotation |
| JetsPart | 1280x720 | 100 | 25fps | Overall convergent-divergent |
Table 2. embodiment one is relative to the BD bit-rate performance of VC-0.2
Claims (10)
1. an inter-frame prediction method in hybrid video coding standard, described Forecasting Methodology is for describing the amoeboid movement existed in video sequence, and described Forecasting Methodology is used in merge pattern, skip pattern or inter pattern;
It is characterized in that, the implementation procedure of described Forecasting Methodology is:
Step one: the movable information obtaining several adjacent coded blocks of surrounding of present encoding block, present encoding block is of a size of W*H, and W is the wide of present encoding block, and H is the height of present encoding block; Described movable information comprises reference key and motion vector; Around several adjacent coded blocks are called contiguous encoding block;
Step 2: the reference key obtaining each division unit in present encoding block according to the reference key of the contiguous encoding block of step one acquisition;
Step 3: process according to the motion vector of reference key to contiguous encoding block of each division unit in the present encoding block that reference key and the step 2 of the contiguous encoding block of step one acquisition obtain, obtain the motion vector of each division unit in present encoding block.
2. inter-frame prediction method in hybrid video coding standard according to claim 1, is characterized in that:
In step one, described contiguous encoding block is the contiguous block of the contiguous block of four corner location of present encoding block or four corner location of present encoding block and center position; Described contiguous encoding block is arranged in the spatial domain contiguous block of present frame or is arranged in the time domain contiguous block of time domain reference frame;
In step 2 and step 3, the size of division unit is w*h, wherein: 1≤w≤W, 1≤h≤H;
The principle obtaining the reference key of each division unit in present encoding block described in step 2 is:
The maximum reference key of the access times of contiguous encoding block obtained in selecting step one as object reference index,
Or for each contiguous encoding block specifies weights, added up by the weights of the contiguous encoding block with same reference index, the reference key of the maximum weight of the contiguous encoding block obtained in selecting step one is as object reference index;
Described in step 3, the motion vector of contiguous encoding block is processed, its process is: before the motion vector interpolation of use contiguous block goes out the motion vector of target division unit in current block, carry out preliminary treatment to the motion vector of contiguous block, if namely the object reference index that obtains with this claim of the reference key of contiguous block is different, just the motion vector of contiguous block zoomed to the motion vector corresponding to object reference index according to time domain distance.
3. inter-frame prediction method in hybrid video coding standard according to claim 2, is characterized in that: in step 3, and in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is:
Described motion vector is calculated by bilinear interpolation model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in claim 2; Preliminary treatment is carried out to choosing the motion vector of contiguous block according to claim 2, then just can according to formula
calculate the motion vector of current block target division unit, in formula, B
mrepresent present encoding block, U
xrepresent certain division unit in current block, d
m(U
x) represent the motion vector MV of target division unit in current block, d
m(MP
k) represent B
mthe MV of a kth contiguous block, interpolation kernel φ
m,kx () depends on B
ma kth contiguous block to U
xcontribution, can be determined by the position of target division unit relative to each control MP.
4. inter-frame prediction method in hybrid video coding standard according to claim 2, it is characterized in that: in step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by the affine model of six parameters, its computational process is: choose three contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in claim 2; According to claim 2, mode carries out preliminary treatment to choosing the motion vector of contiguous block, utilizes the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate six affine transformation coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6be affine transformation coefficient, at least need the motion vector of three contiguous blocks could determine described affine transformation coefficient.
5. inter-frame prediction method in hybrid video coding standard according to claim 2, it is characterized in that: in step 3, the implementation procedure obtaining the motion vector of each division unit in present encoding block is: calculated by the affine model of four parameters, its computational process is: choose two contiguous blocks in the some contiguous blocks from step one, and their movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in claim 2; According to claim 2, mode carries out preliminary treatment to choosing the motion vector of contiguous block, utilizes the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate four affine transformation coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
0, a
1, a
2, a
3be affine transformation coefficient, at least need the motion vector of two contiguous blocks could determine described affine transformation coefficient.
6. inter-frame prediction method in hybrid video coding standard according to claim 2, it is characterized in that: in step 3, in described acquisition present encoding block, the implementation procedure of the motion vector of each division unit is: calculated by perspective projection model, its computational process is: choose four contiguous blocks in the some contiguous blocks from step one, and its movable information all must exist and have at least one to choose the movable information of contiguous block and other to choose the movable information of contiguous block different; The reference key of selected contiguous block is utilized to obtain the object reference index of each division unit in current block according to mode described in claim 2; According to claim 2, mode carries out preliminary treatment to choosing the motion vector of contiguous block, utilizes the position coordinates and motion vector thereof of choosing contiguous block according to formula
Calculate eight perspective projection coefficients, and then corresponding motion vector (vx, vy) can be tried to achieve according to the coordinate of pixel each under same coordinate system in current block; In formula, (x, y) is the coordinate of certain pixel in present frame, and (x', y') is the coordinate of this pixel corresponding points in reference frame, a
1, a
2, a
3, a
4, a
5, a
6, a
7, a
8be perspective projection coefficient, at least need the motion vector of four contiguous blocks could determine described perspective projection coefficient.
7. inter-frame prediction method in hybrid video coding standard according to claim 2, it is characterized in that: in step 3, being embodied as of motion vector of each division unit in described acquisition present encoding block: for current block selects an optimum forecast model in described bilinear interpolation model, perspective projection model, six parameter affine models, four-parameter equation.
8. inter-frame prediction method in hybrid video coding standard according to claim 7, it is characterized in that: the selection course of optimum forecast model is: in the some contiguous blocks obtained from step one, choose four contiguous blocks, the parameter of bilinear interpolation model is calculated according to its relative position, four contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of perspective projection model is calculated according to its relative position and movable information, three contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of six parameter affine models is calculated according to its relative position and movable information, two contiguous blocks are chosen in the some contiguous blocks obtained from step one, the parameter of four-parameter equation is calculated according to its relative position and movable information, then in the set of all effective forecast model formations, select an optimum forecast model or form in the random subset of set at all effective forecast models to select an optimum forecast model, the motion vector of each division unit in present encoding block can be obtained after determining forecast model,
The motion vector of each division unit in described acquisition current block, when selecting optimum forecast model, can be candidate prediction model sequence in candidate prediction model set according to the estimated performance of forecast model or utilization rate.
9. inter-frame prediction method in hybrid video coding standard according to claim 7 or 8, is characterized in that: the decision criteria of selection optimum prediction model is: adopt least mean-square error, minimal error quadratic sum, minimum Hadamard error or rate-distortion optimization criterion.
10. inter-frame prediction method in hybrid video coding standard according to claim 3,4,5,6 or 7, is characterized in that: be first divided into N part to acquisition present encoding block, is the corresponding application of every part a kind of forecast model wherein, wherein N >=1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510415613.6A CN104935938B (en) | 2015-07-15 | 2015-07-15 | Inter-frame prediction method in a kind of hybrid video coding standard |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510415613.6A CN104935938B (en) | 2015-07-15 | 2015-07-15 | Inter-frame prediction method in a kind of hybrid video coding standard |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104935938A true CN104935938A (en) | 2015-09-23 |
| CN104935938B CN104935938B (en) | 2018-03-30 |
Family
ID=54122856
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510415613.6A Active CN104935938B (en) | 2015-07-15 | 2015-07-15 | Inter-frame prediction method in a kind of hybrid video coding standard |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104935938B (en) |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017148345A1 (en) * | 2016-03-01 | 2017-09-08 | Mediatek Inc. | Method and apparatus of video coding with affine motion compensation |
| CN107181952A (en) * | 2016-03-10 | 2017-09-19 | 北京大学 | Video encoding/decoding method and device |
| CN107360419A (en) * | 2017-07-18 | 2017-11-17 | 成都图必优科技有限公司 | A kind of motion forward sight video interprediction encoding method based on perspective model |
| CN107534770A (en) * | 2015-03-10 | 2018-01-02 | 华为技术有限公司 | Image prediction method and relevant device |
| CN108886619A (en) * | 2016-01-07 | 2018-11-23 | 联发科技股份有限公司 | The method and device that affine merging patterns for video coding and decoding system are predicted |
| CN109155855A (en) * | 2016-05-16 | 2019-01-04 | 高通股份有限公司 | Affine motion for video coding is predicted |
| CN109510991A (en) * | 2017-09-15 | 2019-03-22 | 浙江大学 | A kind of motion vector deriving method and device |
| WO2019072187A1 (en) * | 2017-10-13 | 2019-04-18 | Huawei Technologies Co., Ltd. | Pruning of motion model candidate list for inter-prediction |
| CN109862369A (en) * | 2018-12-28 | 2019-06-07 | 杭州海康威视数字技术股份有限公司 | A kind of decoding method and its equipment |
| CN110536135A (en) * | 2018-05-25 | 2019-12-03 | 腾讯美国有限责任公司 | Method and apparatus for coding and decoding video |
| CN110677675A (en) * | 2018-07-01 | 2020-01-10 | 北京字节跳动网络技术有限公司 | Efficient affine Merge motion vector derivation |
| WO2020010741A1 (en) * | 2018-07-11 | 2020-01-16 | 华为技术有限公司 | Video encoder, video decoder, and corresponding method |
| WO2020042604A1 (en) * | 2018-08-27 | 2020-03-05 | 华为技术有限公司 | Video encoder, video decoder and corresponding method |
| CN110876065A (en) * | 2018-08-29 | 2020-03-10 | 华为技术有限公司 | Construction method of candidate motion information list, and inter-frame prediction method and device |
| CN110933423A (en) * | 2018-09-20 | 2020-03-27 | 杭州海康威视数字技术股份有限公司 | Inter-frame prediction method and device |
| CN111083487A (en) * | 2018-10-22 | 2020-04-28 | 北京字节跳动网络技术有限公司 | Storage of motion information in affine mode |
| CN111327901A (en) * | 2020-03-10 | 2020-06-23 | 北京达佳互联信息技术有限公司 | Video encoding method, video encoding device, storage medium and encoding device |
| CN111357288A (en) * | 2019-01-03 | 2020-06-30 | 深圳市大疆创新科技有限公司 | Video image processing method and device |
| CN111418209A (en) * | 2017-10-05 | 2020-07-14 | 交互数字Vc控股公司 | Method and apparatus for video encoding and video decoding |
| WO2020169109A1 (en) * | 2019-02-22 | 2020-08-27 | Beijing Bytedance Network Technology Co., Ltd. | Sub-table for history-based affine mode |
| CN112055205A (en) * | 2017-12-12 | 2020-12-08 | 华为技术有限公司 | Inter-frame prediction method and device of video data |
| US20210266584A1 (en) | 2018-11-16 | 2021-08-26 | Beijing Bytedance Network Technology Co., Ltd. | Usage for history-based affine parameters |
| CN113455001A (en) * | 2018-12-27 | 2021-09-28 | 英迪股份有限公司 | Video encoding/decoding method and apparatus |
| CN113475081A (en) * | 2018-12-05 | 2021-10-01 | 高通股份有限公司 | Triangle motion information for video coding |
| US11206422B2 (en) | 2019-01-03 | 2021-12-21 | SZ DJI Technology Co., Ltd. | Video image processing method and device |
| CN113841408A (en) * | 2019-05-17 | 2021-12-24 | 北京字节跳动网络技术有限公司 | Motion information determination and storage for video processing |
| CN114071147A (en) * | 2020-07-29 | 2022-02-18 | 四川大学 | VVC motion compensation method based on bilinear model |
| CN115243048A (en) * | 2018-12-07 | 2022-10-25 | 华为技术有限公司 | Video image decoding and encoding method and device |
| US11895319B2 (en) | 2018-08-28 | 2024-02-06 | Huawei Technologies Co., Ltd. | Method for constructing candidate motion information list, inter prediction method, and apparatus |
| CN111083487B (en) * | 2018-10-22 | 2024-05-14 | 北京字节跳动网络技术有限公司 | Storage of affine mode motion information |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1359767A1 (en) * | 2002-05-03 | 2003-11-05 | Lg Electronics Inc. | Method for coding moving picture |
| CN104539967A (en) * | 2015-01-15 | 2015-04-22 | 哈尔滨工业大学 | Inter-frame prediction method in mixed video coding standard |
-
2015
- 2015-07-15 CN CN201510415613.6A patent/CN104935938B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1359767A1 (en) * | 2002-05-03 | 2003-11-05 | Lg Electronics Inc. | Method for coding moving picture |
| CN104539967A (en) * | 2015-01-15 | 2015-04-22 | 哈尔滨工业大学 | Inter-frame prediction method in mixed video coding standard |
Non-Patent Citations (4)
| Title |
|---|
| HIROHISA JOZAWA ET AL: "Two-stage motion compensation using adaptive global mc and local affine mc", 《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》 * |
| ROMAN C ET AL: "Affine motion prediction based on translational motion vectors", 《IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY》 * |
| 姚英彪等: "基于图像空间相关性与纹理的HEVC块划分快速算法", 《电信科学》 * |
| 张娜: "下一代视频编码标准中MERGE模式相关技术的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (59)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10404993B2 (en) | 2015-03-10 | 2019-09-03 | Huawei Technologies Co., Ltd. | Picture prediction method and related apparatus |
| US10659803B2 (en) | 2015-03-10 | 2020-05-19 | Huawei Technologies Co., Ltd. | Picture prediction method and related apparatus |
| CN107534770A (en) * | 2015-03-10 | 2018-01-02 | 华为技术有限公司 | Image prediction method and relevant device |
| US11178419B2 (en) | 2015-03-10 | 2021-11-16 | Huawei Technologies Co., Ltd. | Picture prediction method and related apparatus |
| CN107534770B (en) * | 2015-03-10 | 2019-11-05 | 华为技术有限公司 | Image prediction method and relevant device |
| CN108886619A (en) * | 2016-01-07 | 2018-11-23 | 联发科技股份有限公司 | The method and device that affine merging patterns for video coding and decoding system are predicted |
| WO2017148345A1 (en) * | 2016-03-01 | 2017-09-08 | Mediatek Inc. | Method and apparatus of video coding with affine motion compensation |
| CN107181952A (en) * | 2016-03-10 | 2017-09-19 | 北京大学 | Video encoding/decoding method and device |
| CN107181952B (en) * | 2016-03-10 | 2019-11-08 | 北京大学 | Video encoding/decoding method and device |
| CN109155855A (en) * | 2016-05-16 | 2019-01-04 | 高通股份有限公司 | Affine motion for video coding is predicted |
| US11503324B2 (en) | 2016-05-16 | 2022-11-15 | Qualcomm Incorporated | Affine motion prediction for video coding |
| CN107360419A (en) * | 2017-07-18 | 2017-11-17 | 成都图必优科技有限公司 | A kind of motion forward sight video interprediction encoding method based on perspective model |
| CN109510991A (en) * | 2017-09-15 | 2019-03-22 | 浙江大学 | A kind of motion vector deriving method and device |
| CN111418209B (en) * | 2017-10-05 | 2023-10-24 | 交互数字Vc控股公司 | Method and apparatus for video encoding and video decoding |
| CN111418209A (en) * | 2017-10-05 | 2020-07-14 | 交互数字Vc控股公司 | Method and apparatus for video encoding and video decoding |
| WO2019072187A1 (en) * | 2017-10-13 | 2019-04-18 | Huawei Technologies Co., Ltd. | Pruning of motion model candidate list for inter-prediction |
| TWI786790B (en) * | 2017-12-12 | 2022-12-11 | 大陸商華為技術有限公司 | Method and apparatus of frame inter prediction of video data |
| US11363274B2 (en) | 2017-12-12 | 2022-06-14 | Huawei Technologies Co., Ltd. | Video data inter prediction method and apparatus |
| CN112055205B (en) * | 2017-12-12 | 2021-08-03 | 华为技术有限公司 | Inter-frame prediction method and device of video data, video codec and storage medium |
| CN112055205A (en) * | 2017-12-12 | 2020-12-08 | 华为技术有限公司 | Inter-frame prediction method and device of video data |
| CN110536135A (en) * | 2018-05-25 | 2019-12-03 | 腾讯美国有限责任公司 | Method and apparatus for coding and decoding video |
| CN110536135B (en) * | 2018-05-25 | 2021-11-05 | 腾讯美国有限责任公司 | Method and apparatus for video encoding and decoding |
| CN110677675A (en) * | 2018-07-01 | 2020-01-10 | 北京字节跳动网络技术有限公司 | Efficient affine Merge motion vector derivation |
| CN110677675B (en) * | 2018-07-01 | 2022-02-08 | 北京字节跳动网络技术有限公司 | Method, device and storage medium for efficient affine Merge motion vector derivation |
| CN112400318B (en) * | 2018-07-11 | 2024-04-12 | 华为技术有限公司 | Video encoder, video decoder and corresponding methods |
| US11575928B2 (en) | 2018-07-11 | 2023-02-07 | Huawei Technologies Co., Ltd. | Video encoder, video decoder, and corresponding method |
| CN112400318A (en) * | 2018-07-11 | 2021-02-23 | 华为技术有限公司 | Video encoder, video decoder and corresponding methods |
| WO2020010741A1 (en) * | 2018-07-11 | 2020-01-16 | 华为技术有限公司 | Video encoder, video decoder, and corresponding method |
| WO2020042604A1 (en) * | 2018-08-27 | 2020-03-05 | 华为技术有限公司 | Video encoder, video decoder and corresponding method |
| US11895319B2 (en) | 2018-08-28 | 2024-02-06 | Huawei Technologies Co., Ltd. | Method for constructing candidate motion information list, inter prediction method, and apparatus |
| CN110876065A (en) * | 2018-08-29 | 2020-03-10 | 华为技术有限公司 | Construction method of candidate motion information list, and inter-frame prediction method and device |
| CN110933423A (en) * | 2018-09-20 | 2020-03-27 | 杭州海康威视数字技术股份有限公司 | Inter-frame prediction method and device |
| CN110933423B (en) * | 2018-09-20 | 2022-03-25 | 杭州海康威视数字技术股份有限公司 | Inter-frame prediction method and device |
| CN111083487B (en) * | 2018-10-22 | 2024-05-14 | 北京字节跳动网络技术有限公司 | Storage of affine mode motion information |
| CN111083487A (en) * | 2018-10-22 | 2020-04-28 | 北京字节跳动网络技术有限公司 | Storage of motion information in affine mode |
| US11985340B2 (en) | 2018-11-16 | 2024-05-14 | Beijing Bytedance Network Technology Co., Ltd | History-based affine parameters inheritance |
| US11856211B2 (en) | 2018-11-16 | 2023-12-26 | Beijing Bytedance Network Technology Co., Ltd | Usage for history-based affine parameters |
| US11902550B2 (en) | 2018-11-16 | 2024-02-13 | Beijing Bytedance Network Technology Co., Ltd | Pruning method for history-based affine parameters |
| US20210266584A1 (en) | 2018-11-16 | 2021-08-26 | Beijing Bytedance Network Technology Co., Ltd. | Usage for history-based affine parameters |
| CN113475081B (en) * | 2018-12-05 | 2024-04-19 | 高通股份有限公司 | Triangle motion information for video coding |
| CN113475081A (en) * | 2018-12-05 | 2021-10-01 | 高通股份有限公司 | Triangle motion information for video coding |
| CN115243048A (en) * | 2018-12-07 | 2022-10-25 | 华为技术有限公司 | Video image decoding and encoding method and device |
| US11758130B2 (en) | 2018-12-07 | 2023-09-12 | Huawei Technologies Co., Ltd. | Video picture decoding and encoding method and apparatus |
| CN113455001B (en) * | 2018-12-27 | 2024-04-09 | 英迪股份有限公司 | Video encoding/decoding method and apparatus |
| CN113455001A (en) * | 2018-12-27 | 2021-09-28 | 英迪股份有限公司 | Video encoding/decoding method and apparatus |
| CN109862368B (en) * | 2018-12-28 | 2020-02-11 | 杭州海康威视数字技术股份有限公司 | Coding and decoding method and equipment thereof |
| CN109862368A (en) * | 2018-12-28 | 2019-06-07 | 杭州海康威视数字技术股份有限公司 | A kind of decoding method and its equipment |
| US11889072B2 (en) | 2018-12-28 | 2024-01-30 | Hangzhou Hikvision Digital Technology Co., Ltd. | Video encoding and decoding |
| CN109862369A (en) * | 2018-12-28 | 2019-06-07 | 杭州海康威视数字技术股份有限公司 | A kind of decoding method and its equipment |
| US11743482B2 (en) | 2019-01-03 | 2023-08-29 | SZ DJI Technology Co., Ltd. | Video image processing method and device |
| US11689736B2 (en) | 2019-01-03 | 2023-06-27 | SZ DJI Technology Co., Ltd. | Video image processing method and device |
| US11206422B2 (en) | 2019-01-03 | 2021-12-21 | SZ DJI Technology Co., Ltd. | Video image processing method and device |
| CN111357288A (en) * | 2019-01-03 | 2020-06-30 | 深圳市大疆创新科技有限公司 | Video image processing method and device |
| US11863784B2 (en) | 2019-02-22 | 2024-01-02 | Beijing Bytedance Network Technology Co., Ltd | Sub-table for history-based affine mode |
| WO2020169109A1 (en) * | 2019-02-22 | 2020-08-27 | Beijing Bytedance Network Technology Co., Ltd. | Sub-table for history-based affine mode |
| CN113841408B (en) * | 2019-05-17 | 2024-02-06 | 北京字节跳动网络技术有限公司 | Motion information determination and storage for video processing |
| CN113841408A (en) * | 2019-05-17 | 2021-12-24 | 北京字节跳动网络技术有限公司 | Motion information determination and storage for video processing |
| CN111327901A (en) * | 2020-03-10 | 2020-06-23 | 北京达佳互联信息技术有限公司 | Video encoding method, video encoding device, storage medium and encoding device |
| CN114071147A (en) * | 2020-07-29 | 2022-02-18 | 四川大学 | VVC motion compensation method based on bilinear model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104935938B (en) | 2018-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104935938B (en) | Inter-frame prediction method in a kind of hybrid video coding standard | |
| CN111385569B (en) | Coding and decoding method and equipment thereof | |
| KR101633294B1 (en) | Methods For Encoding/Decoding High Definition Image And Apparatuses For Performing The Same | |
| JP2023164435A (en) | Picture prediction method and related apparatus | |
| CN110519600B (en) | Intra-frame and inter-frame joint prediction method and device, coder and decoder and storage device | |
| CN103597831B (en) | Direction intra-frame prediction method and its device | |
| CN101385356B (en) | Process for coding images using intra prediction mode | |
| CN108781284A (en) | The method and device of coding and decoding video with affine motion compensation | |
| BR122013019016B1 (en) | METHOD OF DECODING AN IMAGE | |
| KR20110112240A (en) | Method and apparatus for video encoding, and method and apparatus for video decoding | |
| KR101579106B1 (en) | Method and device for encoding video image, method and device for decoding video image, and program therefor | |
| CN108419082B (en) | Motion estimation method and device | |
| CN110312130B (en) | Inter-frame prediction and video coding method and device based on triangular mode | |
| US20130136187A1 (en) | Video encoding method, video decoding method, video encoding apparatus, video decoding apparatus, and program thereof | |
| CN105491390B (en) | Intra-frame prediction method in hybrid video coding standard | |
| CN103069803B (en) | Method for video coding, video encoding/decoding method, video coding apparatus, video decoder | |
| CN104702959B (en) | A kind of intra-frame prediction method and system of Video coding | |
| CN102984524B (en) | A kind of video coding-decoding method based on block layer decomposition | |
| CN104918047B (en) | A kind of method and device for removing of bi-directional motion estimation | |
| CN101783956A (en) | Back-prediction forecast method based on spatio-temporal neighbor information | |
| CN106063268A (en) | Method and device for transcoding video data from H.264 to H.265 | |
| Zhang et al. | Motion vector derivation of deformable block | |
| CN102204256B (en) | Image prediction method and system | |
| Ratnottar et al. | Review towards the fast block matching algorithms for video motion estimation | |
| CN104539967A (en) | Inter-frame prediction method in mixed video coding standard |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |