CN109510991A

CN109510991A - A kind of motion vector deriving method and device

Info

Publication number: CN109510991A
Application number: CN201710834419.0A
Authority: CN
Inventors: 虞露; 孙煜程
Original assignee: Zhejiang University ZJU
Current assignee: Zhejiang University ZJU
Priority date: 2017-09-15
Filing date: 2017-09-15
Publication date: 2019-03-22
Anticipated expiration: 2037-09-15
Also published as: CN109510991B

Abstract

The present invention provides a kind of motion vector deriving method and device, include: it is similar according to triangle, export the motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates of current decoding unit or current decoding unit intra sub-block using the motion information of current decoding unit periphery decoding region.The motion vector deriving method and device can make full use of the motion information of the adjacent peripheral of current decoding unit decoding region, improve the accuracy derived from motion vector in the case where interframe is common and special pattern, improve encoding-decoding efficiency.

Description

A kind of motion vector deriving method and device

Technical field

The present invention relates to video processing technique more particularly to a kind of motion vector deriving methods and device.

Background technique

In video coding and decoding technology, eliminated by the prediction using time-domain and spatial domain video information in the time and Redundancy spatially.Inter-frame prediction techniques are a kind of technologies for being commonly utilized in coding and decoding video field.By motion compensation, use Information in decoded frame predicts the information of present frame.It needs to transmit many sides in movement compensation process, in video code flow Information, decoding end reconstruct the Pixel Information of present frame using these side informations and transformed residual error.In motion vector In the transmission process of information, by reasonable method of motion vector prediction, the transmission code rate of the effective compression movement information of energy.Fortune The motion vector that dynamic vector export mainly carries out current inter prediction unit using the Space-time domain of current inter prediction unit is pre- Measured value (MVP).To only need to transmit the final motion vector of current inter prediction unit and the difference of MVP in practical code stream (MVD) or not the motion vector MV that MVD directly uses derived MVP as current decoding unit is passed, it is entire most not need transmission Whole motion vector.Under special circumstances, the reference frame information of current inter prediction unit can also be kept and ginseng corresponding to MVP It examines frame information to be consistent, that is, does not need the reference frame information of the current inter prediction unit of additional transmissions.

AMVP (advanced motion-vector prediction) technology is by building motion vector candidates list, by transmitting in code stream The serial number of the motion vector predictor or motion vector chosen, decoding end are sweared according to serial number and by the movement of same rule building Amount candidate list comes derive motion vector or motion vector predictor.AMVP technical application is referred to as Merge in a special mode Mode.Merge mode in derive motion vector, not merely with adjacent block in motion vector information, also utilize simultaneously The reference frame information of adjacent block.In the candidate list of Merge mode construction, the adjacent block according to priority sequentially obtained is housed Motion information, including motion vector information and reference frame information.That is, Merge mode is in the same of derive motion vector When be also derived reference frame corresponding with the motion vector.In Merge mode, current block not only want and select by motion vector Adjacent block it is identical, reference frame also wants identical.The thought of Merge mode is that decoding block belongs to same put down for current block and periphery Movable model, i.e. motion vector MV are consistent.

AMVP and Merge the problem is that, current block only depends on the predicted value of periphery block export motion information. If AMVP and Merge can not Accurate Predictions when the motion model of current block and periphery block is not translational motion model The motion vector of current block.

STMVP (Space-time domain motion-vector prediction) technology is by the motion vector (if present) of the top block of current block, a left side The motion vector (if present) of side block and the TMVP (time domain motion vector predictor) of current block zoom to current block reference frame The first frame of list.This most three motion vector is averaged, to obtain the new motion vector predictor of current block. The problem of STMVP technology, is, although multiple periphery block joint one new MVP of export are utilized, using directly average Method there is no accurate physical significance, i.e., do not make full use of the motion relevance between adjacent block.

Under perspective projection, the movement of rigid body in three dimensions, the projection on imaging plane is rendered as translating, rotation Turn, scaling.Affine motion model is to consider translation, rotation, the motion information expression model of scaling.General affine motion mould There are six parameters for type tool:

(x, y) is the point of present frame in above formula, and (x ', y ') is the point of reference frame, and a, b, c, d, e, f are affine motion models Six parameters.In order to export this six parameters, need 3 groups known to (x, y) and (x ', y ') right, three groups of equation groups of simultaneous, from And a is calculated, and b, c, d, e, the value of six parameters of f.

Under the simplified model of affine motion model, several zonules for the sizes such as interframe prediction block is divided into, Mei Ge little It (is not the possible difference of movement velocity of each pixel, this is a kind of that region (i.e. sub-block) interior movement velocity, which is consistent, The method for reducing the delineation gradual change movement velocity of precision), and the motion compensation model of each zonule is still plane translation mould (image block only translates type in the plane of delineation, does not change shape, size, therefore to the description of sub-block motion still parameterisable For a motion vector).When the rotary shaft of the rotary motion of object is perpendicular to the plane of delineation, any two points of affine object Scaling ratio is consistent (corner dimension that any two straight lines are constituted remains unchanged).A, b, c, d under above-mentioned limitation, The affine motion of six parameters of e, f degenerate be four parameters affine motion model, i.e. a, b exist between tetra- parameters of c, d certain Relationship, only need to be exported a, b, c, tetra- parameters of d as follows with two parameters.

S is the elongated of current block, because current block is rectangular, i.e. S × S in existing affine model.Referring to Figure 10, v_x0For v₀Cross stream component, v_y0For v₀Longitudinal component；v_x1For v₁Cross stream component, v_y1For v₁Longitudinal component.

This special affine motion need to carry out the export of affine motion vector field using two control points.Two control points Set the top left corner pixel point position of generally current interframe prediction block, the upper right corner pixel position of current interframe prediction block.

The region of affine motion is generally in current interframe prediction block, and the motor pattern of decoding block is unrelated with periphery.Before It is that decoding block belongs to a translational Motion Model, i.e. motion vector and reference frame information for current block and periphery that Merge mode is mentioned in face Unanimously.The thought of the Merge mode of affine motion is that decoding block belongs to the same affine motion mould for current block and periphery Type shares affine parameter.And Merge mode is that decoding block belongs to approximately the same plane translational Motion Model for current block and periphery, is shared One motion vector.

Summary of the invention

It is an object of the invention to the motion informations of the periphery decoding region using current decoding unit, export current solution Code unit or current decoding unit intra sub-block motion vector predictor or motion vector or motion vector predictor candidate or Motion vector candidates.

The present invention is based on the affine motion model of four parameters, that is, support the translation of object, scaling and rotary shaft perpendicular to The rotary motion of the plane of delineation.Mentality of designing of the present invention is: the prior art takes the point inside current decoding unit as control Point position, limits to a certain extent through decoding region and exports the accuracy of control point motion vector.And it is of the invention Affine motion model control point is not the point inside current decoding unit but the point except current decoding unit, this is convenient for The motion vector value at accurate control point is exported using decoded region.

The present invention proposes a kind of motion vector deriving method comprising:

According to the similitude of triangle ABC and triangle A ' B ' C ', the periphery decoding region of current decoding unit is utilized Motion vector MV1 and motion vector MV2, MV1 is not equal to MV2, exports the position of the corresponding points C ' of C, wherein A, B is current Point outside decoding unit determines that the terminating point of motion vector MV1 is movement with B for A ' by the starting point of motion vector MV1 of A The starting point of vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is inside current decoding unit or current decoding unit The representative locations of sub-block；

Vector CC ' is calculated by position C and position C ', vector CC ' output is that current decoding unit or current decoding are single The motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates of first intra sub-block.

Wherein so-called sub-block, is the further division of current decoding unit, and sub-block is not equal to current decoding unit.

Preferably, the motion vector deriving method, further includes a reasonableness test step, the reasonableness test Step is according to following criterion:

The ratio of the length of the length and vector A ' B ' of the vector AB is without departing from [k, l] range, 0≤k≤l≤10；

The corner dimension of the vector AB and vector A ' B ' are less than or equal to 90 degree without departing from θ range, θ；

If meeting at least one above criterion, determine that the vector CC ' output is effective.

Preferably, the motion vector deriving method further includes a kind of following situation:

From the left side of current decoding unit, decoding region obtains the motion vector MV1, from current decoding unit Decoding region obtains the motion vector MV2 for top；

From the left side of current decoding unit, decoding region obtains the motion vector MV2, from current decoding unit Decoding region obtains the motion vector MV1 for top.

Preferably, the motion vector deriving method further include by current decoding unit current decoding in the picture Motion vector MV1 and motion vector described in obtaining are searched in decoding region on decoded reference picture on periphery outside unit MV2。

Preferably, the current decoding unit intra sub-block is at least two, each current decoding unit intra sub-block Representative locations it is respectively independent, the vector CC ' that each current decoding unit intra sub-block is calculated by position C and position C ' Also respectively independent.

Another object of the present invention, which also resides in, proposes a kind of motion vector guiding device comprising,

Position export module: according to the similitude of triangle ABC and triangle A ' B ' C ', the week of current decoding unit is utilized The motion vector MV1 of decoding region and motion vector MV2, MV1 export the position of the corresponding points C ' of C not equal to MV2 on side, Middle A, B are the point outside current decoding unit, determine that the terminating point of motion vector MV1 is by the starting point of motion vector MV1 of A A ' determines the terminating point of motion vector MV2 by the starting point of motion vector MV2 of B as B ', and C is for current decoding unit or currently The representative locations of decoding unit intra sub-block；

Motion vector generation module: vector CC ' is calculated by position C and position C ', vector CC ' output is current decoding The motion vector predictor or motion vector or motion vector predictor candidate or fortune of unit or current decoding unit intra sub-block Dynamic vector is candidate.

Preferably, the motion vector guiding device includes adding reasonability inspection before the position export module Module is tested, the reasonableness test module is according to following criterion:

If meeting at least one above criterion, the arrow of the position export module and motion vector generation module is determined The output for measuring CC ' is effective.

Preferably, the motion vector guiding device further includes the addition movement arrow before the position export module Amount obtains module, and it includes a kind of situation below that the motion vector, which obtains module:

Preferably, the motion vector guiding device further include: add movement before the position export module Vector search export module, the motion-vector search export module is for current in the image where current decoding unit The motion vector MV1 described in obtaining and movement are searched in decoding region on decoded reference picture on periphery outside decoding unit Vector MV2.

Compared to the prior art, whether existing affine motion needs are transmitted in the code stream of current decoding unit using imitative The grammer of model is penetrated, but the present invention can be according only to the code stream except current decoding unit code stream, by examining affine model Physical reasonableness, for example whether scaling reasonable, and whether rotation angle reasonable, to determine the movement arrow of current decoding unit It measures predicted value or motion vector or motion vector predictor candidate or whether motion vector candidates is exported by affine model.

In addition, the present invention has also carried out certain limitation to the control point of affine motion model: on the one hand, control point difference Top and left area from current decoding unit, can guarantee to a certain extent control point MV to current decoding unit or The influence degree of current decoding unit intra sub-block；On the other hand, re-searching for for motion vector is carried out with decoding region, it can To guarantee the accuracy of control point motion vector MV, so that it is guaranteed that accuracy derived from affine motion vector.

With affine motion model, current decoding unit is carried out derived from affine motion vector with sub-block grade When, each sub-block should share same set of control point parameter, i.e. the movement arrow of Compliance control point position and control point position MV1 and MV2 is measured to guarantee the affine motion consistency of entire current decoding block.

Detailed description of the invention

In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with It obtains other drawings based on these drawings.

Fig. 1 is a kind of current decoding unit of interframe provided in an embodiment of the present invention and its adjacent block schematic diagram；

Fig. 2 is a kind of schematic diagram using motion vector MV1 and MV2 derive motion vector provided in an embodiment of the present invention；

Fig. 3 is a kind of showing for motion vector that sub-block is exported using motion vector MV1 and MV2 provided in an embodiment of the present invention It is intended to；

Fig. 4 is that the current decoding unit of a kind of interframe provided in an embodiment of the present invention divides sub-block and its adjacent block schematic diagram；

Fig. 5 (a) is a kind of schematic diagram of motion vector guiding device provided in an embodiment of the present invention；

Fig. 5 (b) is a kind of schematic diagram of sub-block motion vectors guiding device provided in an embodiment of the present invention；

Fig. 6 (a) is a kind of schematic diagram of motion vector guiding device provided in an embodiment of the present invention；

Fig. 6 (b) is a kind of schematic diagram of sub-block motion vectors guiding device provided in an embodiment of the present invention；

Fig. 7 (a) is a kind of schematic diagram of motion vector guiding device provided in an embodiment of the present invention；

Fig. 7 (b) is a kind of schematic diagram of sub-block motion vectors guiding device provided in an embodiment of the present invention；

Fig. 8 (a) is a kind of schematic diagram of motion vector guiding device provided in an embodiment of the present invention；

Fig. 8 (b) is a kind of schematic diagram of sub-block motion vectors guiding device provided in an embodiment of the present invention；

Fig. 9 is a kind of schematic diagram of sub-block motion vectors guiding device provided in an embodiment of the present invention；

Figure 10 is the schematic diagram of four parameter affine motion models in background of invention.

Specific embodiment

In Video coding, video data can be divided into different size of M × N (value of M, N are generally 2 power) Inter-prediction block of pixels.The block of pixels to be decoded of these M × N is known as current decoding unit.It can be with around current decoding unit There are multiple decoded pixel blocks, decoded pixel block is referred to as periphery decoding block for these, and periphery decoding block region is Periphery decoding region.

In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.

Embodiment 1

A kind of motion vector deriving method that this example provides, specifically includes:

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current decoding unit.

Specifically, a kind of available deriving method: as shown in Figure 1, decoding block has A1 on the periphery of current decoding unit, A2 ... ..., An, B1, B2 ... ... Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k ≤n+m.From the acquisition motion vector MV1 and motion vector MV2 in k interframe prediction block.Interframe prediction block where A takes MV1 Center, the center of prediction block, C take the center of current decoding unit, A, B, the physics of location of C where B takes MV2 Meaning is not bound.

As shown in Fig. 2, triangle ABC and triangle A ' B ' C ' meet following relationship:

ΔABC∽ΔA′B′C′

That is, in the similarity relation of triangle ABC and triangle A ' B ' C ', A corresponding A ', B corresponds to B ', and C corresponds to C '.Vector AA ' is motion vector MV1, and vector BB ' is motion vector MV2, and needing derived motion vector MV3 is vector CC '.Assuming that θ₁It is The angle of vector AB and vector A ' B '；θ′₁It is the angle of vector AC and vector A ' C '；Assuming that θ₂It is the folder of vector BA and vector AC Angle；θ′₂It is the angle of vector B ' A ' and vector A ' C ', then the similitude of triangle ABC and triangle A ' B ' C ' can use following formula It indicates:

If (a₁,b₁) it is vector AB, (a₂,b₂) it is vector A ' B ', (c₁,d₁) it be vector AC is (c₂,d₂) it is vector A ' C ', (e₁,f₁) it is vector CB.The similitude of triangle ABC and triangle A ' B ' C ' is rewritten are as follows:

The value (x ', y ') of the position C ' is obtained by the similitude of above formula triangle ABC and triangle A ' B ' C ':

α, β are obtained by following formula₁,β₂:

Function g (k, l) and function h (k, l) are obtained by following formula:

Wherein, (x '₁,y′₁) be point A ' coordinate, (x '₂,y′₂) be point B ' coordinate.

The value (x ', y ') of the position C ' can also be calculated by other processes of equal value with this or with this approximate process It arrives.

Step 2: vector CC ' is calculated by position C and position C ', vector CC ' output is the movement of current decoding unit Vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX.

Specifically, according to the coordinate of the coordinate (x, y) of C and C ' (x ', y '), by the cross stream component of following formula method export MVX The longitudinal component MVXy of MVXx and MVX:

Embodiment 2

Specifically, a kind of available deriving method such as the step of embodiment 1 one shown in.

Embodiment 3

Embodiment 4

Embodiment 5

Embodiment 6

Embodiment 7

Step 1 judges according to the corner dimension of vector AB and vector A ' B ' without departing from θ (θ is less than or equal to 90 degree) range Whether step 2 is carried out.

Specifically, the parallel degree of line segment AB and line segment A ' B ' are examined.The parallel degree of line segment AB and line segment A ' B ' are portrayed The degree of rotation of object, and the rotation between reference frame and present frame be more than to a certain degree after, the affine model of block grade is no longer It is applicable in.So the angle as line segment AB and line segment A ' B ' is less than given threshold value γ, is judged to examining rationally.

Step 2 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current decoding unit.

Step 3: vector CC ' is calculated by position C and position C ', vector CC ' output is the movement of current decoding unit Vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX.

Embodiment 8

Step 1, according to the ratio of the length of vector AB and the length of vector A ' B ' without departing from [k, l] (0≤k≤l≤10) Range is to determine whether carry out step 2.

Specifically, the length ratio size of line segment AB and line segment A ' B ' are detected.Because of the length of line segment AB and line segment A ' B ' Scale features the zoom degree of object, and the scaling between reference frame and present frame be more than to a certain degree after, block grade Affine model is no longer applicable in.So when the length ratio size of line segment AB and line segment A ' B ' is in given range [1/ δ, δ], δ >=1 It is interior, it is judged to examining rationally, carries out step 2.

Embodiment 9

Step 1 obtains information from the code stream of current decoding unit, according to what is obtained in the code stream of current decoding unit Information decides whether to carry out step 2.

Specifically, a kind of method obtaining information from the code stream of current decoding unit is, in the code of current decoding unit A flag bit is transmitted in stream.When flag bit is 1 after decoding, step 2 operation is carried out.

Embodiment 10

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current decoding unit.Its Middle motion vector MV1 and motion vector MV2 both is from the left side, and decoding region or motion vector MV1 and motion vector MV2 come From top decoding region.

Specifically, as shown in Figure 1, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 and motion vector MV2 in prediction block or from obtaining in m top decoding region interframe prediction block Take motion vector MV1 and motion vector MV2.It the position of ABC can position in Example 1,2,3,4,5,6 and other representatives Property position, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 11

Specifically, as shown in Figure 1, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 in prediction block, from the acquisition motion vector MV2 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current decoding unit.The position of ABC is desirable to implement Position and other representative locations in example 1,2,3,4,5,6, wherein the position of AB is outside current decoding unit, A, B, and C The physical significance set is not bound.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 12

Specifically, as shown in Figure 1, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV2 in prediction block, from the acquisition motion vector MV1 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current decoding unit.The position of ABC is desirable to implement Position and other representative locations in example 1,2,3,4,5,6, wherein the position of AB is outside current decoding unit, A, B, and C The physical significance set is not bound.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 13

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current decoding unit.Its Middle motion vector MV1 and motion vector MV2 be as the periphery outside decoding unit current in the image where current decoding unit Decoding region is searched on decoded reference picture and obtains motion vector MV1 and motion vector MV2.

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.The position of ABC can position in Example 1,2,3,4,5,6 and its His representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.Take MV1 institute Periphery decoding block the center region S × W, referred to as regional area.Using the region on the reference frame that MV1 is directed toward With the smallest criterion of SAD or SATD, the export of the value of motion vector MV1 is carried out using MV1 as initial search point；Where taking MV2 Center S × W area, referred to as regional area of periphery decoding block.Be directed toward on reference frame with SAD using the region in MV1 or The smallest criterion of SATD take the flexible value of the time domain of MV2 as the export for the value that initial search point carries out motion vector MV2.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 14

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current decoding unit.Its Middle motion vector MV1 and motion vector MV2 be as the periphery outside decoding unit current in the image where current decoding unit Decoding region, which is searched on decoded reference picture, to be obtained.

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.The position of ABC can position in Example 1,2,3,4,5,6 and its His representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.Take MV1 institute Periphery decoding block the center region S × W, referred to as regional area.Using the region on the reference frame that MV2 is directed toward It take the flexible value of the time domain of MV1 as the export for the value that initial search point carries out motion vector MV1 with the smallest criterion of SAD or SATD； Take center S × W area, referred to as regional area of periphery where MV2 decoding block.It is directed toward and is referred in MV2 using the region With the smallest criterion of SAD or SATD on frame, the export of the value of motion vector MV2 is carried out using MV2 as initial search point.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 15

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.The position of ABC can position in Example 1,2,3,4,5,6 and its His representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.Take MV1 institute Periphery decoding block the center region S × W, referred to as regional area.Using the region present frame a certain reference frame Ref is upper with the smallest criterion of SAD or SATD, is the value that initial search point carries out motion vector MV1 with the flexible value of the time domain of MV1 Export；Take center S × W area, referred to as regional area of periphery where MV2 decoding block.It is being referred to using the region It is the value that initial search point carries out motion vector MV2 with the flexible value of the time domain of MV2 with the smallest criterion of SAD or SATD on frame Ref Export.

Further deriving method such as the step of embodiment 1 one shown in.

Embodiment 16

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current sub-block.

Specifically, a kind of available deriving method: as shown in figure 3, decoding block has A1 on the periphery of current decoding unit, A2 ... ..., An, B1, B2 ... ... Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k ≤n+m.From the acquisition motion vector MV1 and motion vector MV2 in k interframe prediction block.Interframe prediction block where A takes MV1 Center, the center of prediction block, C take the center of current sub-block, A, B, the physical significance of location of C where B takes MV2 It does not bind.

As shown in figure 3, triangle ABC and triangle A ' B ' C ' meet following relationship:

ΔABC∽ΔA′B′C′

α, β are obtained by following formula₁,β₂:

Function g (k, l) and function h (k, l) are obtained by following formula:

Step 2: vector CC ' is calculated by position C and position C ', vector CC ' output is the motion vector of current sub-block Predicted value or motion vector or motion vector predictor candidate or motion vector candidates MVX.

Embodiment 17

Specifically, a kind of available deriving method such as the step of embodiment 16 one shown in.

Embodiment 18

Embodiment 19

Embodiment 20

Embodiment 21

Embodiment 22

Step 1, it is only necessary to the code stream except current sub-block code stream, according to the motion vector MV1 and motion vector The physical significance reasonability of the point position MV2 and A and B point position is to determine whether carry out step 2.

Specifically, the parallel degree of line segment AB and line segment A ' B ' are examined.The parallel degree of line segment AB and line segment A ' B ' are portrayed The degree of rotation of object, and after the rotation to a certain degree between reference frame and present frame, the affine model of block grade is no longer applicable in. So the angle as line segment AB and line segment A ' B ' is less than given threshold value γ, is judged to examining rationally.

Step 2 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current sub-block.

Step 3: vector CC ' is calculated by position C and position C ', vector CC ' output is the motion vector of current sub-block Predicted value or motion vector or motion vector predictor candidate or motion vector candidates MVX.

Embodiment 23

Embodiment 24

Step 1 obtains information from the code stream of current sub-block, is determined according to the information obtained in the code stream of current sub-block Whether step 2 is carried out.

Specifically, a kind of method that information is obtained in code stream from current sub-block is to transmit in the code stream of current sub-block One flag bit.When flag bit is 1 after decoding, step 2 operation is carried out.

Embodiment 25

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current sub-block.Wherein transport Dynamic vector MV1 and motion vector MV2 both is from the left side, and decoding region or motion vector MV1 and motion vector MV2 both are from Side decoding region.

Specifically, as shown in figure 4, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 and motion vector MV2 in prediction block or from obtaining in m top decoding region interframe prediction block Take motion vector MV1 and motion vector MV2.The position of ABC can position in Example 16,17,18,19,20,21 and its His representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 26

Specifically, as shown in figure 4, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 in prediction block, from the acquisition motion vector MV2 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current sub-block.The position of ABC can Example 16, Position and other representative locations in 17,18,19,20,21, wherein the position of AB is outside current decoding unit, A, B, and C The physical significance set is not bound.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 27

Specifically, as shown in figure 4, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV2 in prediction block, from the acquisition motion vector MV1 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current sub-block.The position of ABC can Example 16, Position and other representative locations in 17,18,19,20,21, wherein the position of AB is outside current decoding unit, A, B, and C The physical significance set is not bound.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 28

Step 1 has been solved according to the similitude of triangle ABC and triangle A ' B ' C ' using the periphery of current decoding unit The motion vector MV1 and motion vector MV2, MV1 in code region export the position of the corresponding points C ' of C, wherein A, B are not equal to MV2 Point outside current decoding unit, determines that the terminating point of motion vector MV1 for A ', is with B by the starting point of motion vector MV1 of A The starting point of motion vector MV2 determines that the terminating point of motion vector MV2 is B ', and C is the representative locations of current sub-block.Wherein transport Dynamic vector MV1 and motion vector MV2 has been decoded as the periphery outside decoding unit current in the image where current decoding unit Region, which is searched on decoded reference picture, to be obtained.

Specifically, as shown in figure 4, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.It the position of ABC can position in Example 16,17,18,19,20,21 And other representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.It takes Center S × W area, referred to as regional area of periphery where MV1 decoding block.The reference being directed toward using the region in MV1 With the smallest criterion of SAD or SATD on frame, the export of the value of motion vector MV1 is carried out using MV1 as initial search point；Take MV2 institute Periphery decoding block center S × W area, referred to as regional area.Using the region MV1 be directed toward reference frame on The smallest criterion of SAD or SATD take the flexible value of the time domain of MV2 as the export for the value that initial search point carries out motion vector MV2.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 29

Specifically, as shown in figure 4, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.It the position of ABC can position in Example 16,17,18,19,20,21 And other representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.It takes Center S × W area, referred to as regional area of periphery where MV1 decoding block.The reference being directed toward using the region in MV2 It is the value that initial search point carries out motion vector MV1 with the flexible value of the time domain of MV1 with the smallest criterion of SAD or SATD on frame Export；Take center S × W area, referred to as regional area of periphery where MV2 decoding block.It is directed toward using the region in MV2 With the smallest criterion of SAD or SATD on reference frame, the export of the value of motion vector MV2 is carried out using MV2 as initial search point.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 30

Specifically, as shown in figure 4, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.It the position of ABC can position in Example 16,17,18,19,20,21 And other representative locations, wherein the position of AB is outside current decoding unit, A, B, and the physical significance of location of C is not bound.It takes Center S × W area, referred to as regional area of periphery where MV1 decoding block.Using the region in a certain of present frame It is that initial search point carries out motion vector with the flexible value of the time domain of MV1 with the smallest criterion of SAD or SATD on reference frame Ref The export of the value of MV1；Take the center region S × W of periphery where MV2 decoding block, referred to as regional area.Utilize the area Domain, with the smallest criterion of SAD or SATD, carries out movement arrow using the flexible value of the time domain of MV2 as initial search point on reference frame Ref Measure the export of the value of MV2.

Further deriving method such as the step of embodiment 16 one shown in.

Embodiment 31

A kind of motion vector guiding device that this example provides, as shown in Fig. 5 (a):

Module S101, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, current decoding unit representative locations C point position coordinates；Its output includes: C ' puts position coordinates；

The function of the module are as follows: according to the similitude of triangle ABC and triangle A ' B ' C ', utilize current decoding unit The motion vector MV1 of decoding region and motion vector MV2, MV1 export the position of the corresponding points C ' of C not equal to MV2 on periphery, The terminating point of motion vector MV1 is determined by the starting point of motion vector MV1 of A for A ', using B as the starting point of motion vector MV2 The terminating point for determining motion vector MV2 is B '.

Specifically, a kind of available deriving method: as shown in Fig. 2, triangle ABC and triangle A ' B ' C ' meets such as ShiShimonoseki System:

ΔABC∽ΔA′B′C′

α, β are obtained by following formula₁,β₂:

Function g (k, l) and function h (k, l) are obtained by following formula:

Module S102, input include: C ' point position coordinates；Its output includes motion vector MVX；

The function of the module are as follows: vector CC ' is calculated by position C and position C ', vector CC ' output is single for current decoding The motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX of member.

Embodiment 32

A kind of motion vector guiding device that this example provides, as shown in Fig. 5 (b):

Module S201, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, current sub-block representative locations C point position coordinates；Its output includes: C ' point Position coordinates；

Specifically, a kind of available deriving method is as shown in the module S101 of embodiment 31.

Module S202, input include: C ' point position coordinates；Its output includes motion vector MVX；

The function of the module are as follows: vector CC ' is calculated by position C and position C ', vector CC ' output be current sub-block or The motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX of current decoding unit.

Embodiment 33

A kind of motion vector guiding device that this example provides, as shown in Fig. 6 (a):

Module S301, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit；Its output includes: reasonableness test result；

The function of the module are as follows: according to the corner dimension of vector AB and vector A ' B ' without departing from θ (θ is less than or equal to 90 degree) model It encloses to judge reasonability.

Specifically, the parallel degree of line segment AB and line segment A ' B ' are examined.The parallel degree of line segment AB and line segment A ' B ' are portrayed The degree of rotation of object, and the rotation between reference frame and present frame be more than to a certain degree after, the affine model of block grade is no longer It is applicable in.So the angle as line segment AB and line segment A ' B ' is less than given threshold value γ, it is determined as " reasonable ".

Module S302, input include: reasonableness test as a result, motion vector MV1, motion vector MV2, current decoding are single A point position coordinates outside member, the B point position coordinates outside current decoding unit, current decoding unit representative locations C point position Coordinate；It includes: C ' point position coordinates that it, which is exported,；

The function of the module are as follows: if reasonableness test result is " reasonable ", according to triangle ABC's and triangle A ' B ' C ' Similitude, using the periphery of current decoding unit, the motion vector MV1 of decoding region and motion vector MV2, MV1 are not equal to MV2 exports the position of the corresponding points C ' of C, determines that the terminating point of motion vector MV1 is by the starting point of motion vector MV1 of A A ' determines the terminating point of motion vector MV2 for B ' by the starting point of motion vector MV2 of B.

Module S303, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 34

The function of the module are as follows: according to the ratio of the length of vector AB and the length of vector A ' B ' without departing from [k, l] (0≤k ≤ l≤10) range judges reasonability.

Specifically, the length ratio size of line segment AB and line segment A ' B ' are detected.Because of the length of line segment AB and line segment A ' B ' Scale features the zoom degree of object, and the scaling between reference frame and present frame be more than to a certain degree after, block grade Affine model is no longer applicable in.So when the length ratio size of line segment AB and line segment A ' B ' is in given range [1/ δ, δ], δ >=1 It is interior, it is determined as " reasonable ".

Embodiment 35

A kind of motion vector guiding device that this example provides, as shown in Fig. 6 (b):

Module S401, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit；Its output includes: reasonableness test result；

The function of the module are as follows: the angle according to vector AB and vector A ' B ' is to judge to close without departing from [0 °, 90 °] range Rationality.

Module S402, input include: reasonableness test as a result, motion vector MV1, motion vector MV2, current decoding are single A point position coordinates outside member, the B point position coordinates outside current decoding unit, current sub-block representative locations C point position coordinates； It includes: C ' point position coordinates that it, which is exported,；

Module S403, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 36

Embodiment 37

A kind of motion vector guiding device that this example provides, as shown in Fig. 7 (a):

Module S501, input include: the motion vector of current decoding unit periphery decoding block；It includes: fortune that it, which is exported, Dynamic vector MV1, motion vector MV2；

The function of the module are as follows: decoding region obtains motion vector MV1 and decoding region obtains from top from the left side Motion vector MV2 or from the left side decoding region obtain motion vector MV2 and from top decoding region obtain motion vector MV1。

Specifically, as shown in Figure 1, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 in prediction block, from the acquisition motion vector MV2 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current decoding unit.

Module S502, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, current decoding unit representative locations C point position coordinates；Its output includes: C ' puts position coordinates；

Module S503, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 38

A kind of motion vector guiding device that this example provides, as shown in Fig. 7 (b):

Module S601, input include: the motion vector of current decoding unit periphery decoding block；It includes: fortune that it, which is exported, Dynamic vector MV1, motion vector MV2；

Specifically, as shown in figure 4, decoding block has A1 on the periphery for belonging to left side decoding region of current decoding unit, A2 ... ..., An, decoding block has B1, B2 ... ... Bm on the periphery for belonging to top decoding region of current decoding unit, wherein N, m is the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From n left side decoding region interframe Acquisition motion vector MV1 in prediction block, from the acquisition motion vector MV2 in m top decoding region interframe prediction block.With This improves motion vector MV1 and motion vector MV2 to the influence degree of entire current decoding unit.

Module S602, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, the representative locations C point position coordinates of current sub-block；Its output includes: C ' Point position coordinates；

Module S603, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 39

A kind of motion vector guiding device that this example provides, as shown in Fig. 8 (a):

Module S701, input include: the motion vector of current decoding unit periphery decoding block；It includes: fortune that it, which is exported, Dynamic vector MV1, motion vector MV2；

The function of the module are as follows: as the area decoder of the periphery outside decoding unit current in the image where current decoding unit Domain is searched on decoded reference picture and obtains motion vector MV1 and motion vector MV2.

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.Center S × W area of periphery where MV1 decoding block is taken, referred to as Regional area.Using the region with the smallest criterion of SAD or SATD on the reference frame that MV1 is directed toward, using MV1 as initial search point Carry out the export of the value of motion vector MV1；Take center S × W area, referred to as partial zones of periphery where MV2 decoding block Domain.It is directed toward on reference frame in MV1 with the smallest criterion of SAD or SATD using the region, is search with the flexible value of the time domain of MV2 Starting point carries out the export of the value of motion vector MV2.

Module S702, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, current decoding unit representative locations C point position coordinates；Its output includes: C ' puts position coordinates；

Module S703, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 40

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.Center S × W area of periphery where MV1 decoding block is taken, referred to as Regional area.Using the region with the smallest criterion of SAD or SATD on the reference frame that MV2 is directed toward, with the flexible value of the time domain of MV1 The export of the value of motion vector MV1 is carried out for initial search point；Center S × W area of periphery where MV2 decoding block is taken, Referred to as regional area.It is directed toward on reference frame in MV2 with the smallest criterion of SAD or SATD using the region, is to search for MV2 Initial point carries out the export of the value of motion vector MV2.

Embodiment 41

Specifically, as shown in Figure 1, currently decoding block has A1, A2 ... ..., An, B1, B2 ... ... on the periphery of decoding unit Bm, wherein n, m are the natural number greater than 0.Wherein having k block is interframe prediction block, 0≤k≤n+m.From k interframe prediction block In acquisition motion vector MV1 and motion vector MV2.Center S × W area of periphery where MV1 decoding block is taken, referred to as Regional area.Using the region with the smallest criterion of SAD or SATD on a certain reference frame Ref of present frame, with the time domain of MV1 Flexible value is the export for the value that initial search point carries out motion vector MV1；Take center S × W of periphery where MV2 decoding block Region, referred to as regional area.Using the region with the smallest criterion of SAD or SATD on reference frame Ref, with the time domain of MV2 Flexible value is the export for the value that initial search point carries out motion vector MV2.

Embodiment 42

A kind of motion vector guiding device that this example provides, as shown in Fig. 8 (b):

Module S801, input include: the motion vector of current decoding unit periphery decoding block；It includes: fortune that it, which is exported, Dynamic vector MV1, motion vector MV2；

Module S802, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, current sub-block representative locations C point position coordinates；Its output includes: C ' point Position coordinates；

Module S803, input include: C ' point position coordinates；Its output includes motion vector MVX；

Embodiment 43

Embodiment 44

Embodiment 45

By taking current decoding unit divides four sub-blocks as an example (1,2,3,4 is four sub-blocks):

1	2
		3	4

Step 1, according to the similitude of triangle ABC2 and triangle A ' B ' C2 ', using current decoding unit periphery The motion vector MV1 and motion vector MV2 of decoding region, export the position of the corresponding points C2 ' of C2, wherein A, and B is current decoding Point outside unit determines the terminating point of motion vector MV1 by the starting point of motion vector MV1 of A for A ', using B as motion vector The starting point of MV2 determines that the terminating point of motion vector MV2 is B ', and C2 is the representative locations of the 2nd sub-block.

According to the similitude of triangle ABC3 and triangle A ' B ' C3 ', the periphery area decoder of current decoding unit is utilized The motion vector MV1 and motion vector MV2 in domain, export the position of the corresponding points C3 ' of C3, wherein A, and B is outside current decoding unit Point, determined by the starting point of motion vector MV1 of A the terminating point of motion vector MV1 for A ', using B as motion vector MV2 rise Initial point determines that the terminating point of motion vector MV2 is B ', and C3 is the representative locations of the 3rd sub-block.

Step 2: vector C2C2 ' is calculated by position C2 and position C2 ', vector C2C2 ' output is the 2nd sub-block Motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX2.

Specifically, according to the coordinate (x of C₂,y₂) and C ' coordinate (x₂’,y₂'), by the transverse direction of following formula method export MVX2 The longitudinal component MVX2y of component MVX2x and MVX2:

Vector C3C3 ' is calculated by position C3 and position C3 ', vector C3C3 ' output is the motion vector of the 3rd sub-block Predicted value or motion vector or motion vector predictor candidate or motion vector candidates MVX3.

Specifically, according to the coordinate (x of C₃,y₃) and C ' coordinate (x₃’,y₃'), by the transverse direction of following formula method export MVX1 The longitudinal component MVX1y of component MVX1x and MVX1:

Embodiment 46

1	2
		3	4

Step 1, according to the similitude of triangle ABC1 and triangle A ' B ' C1 ', using current decoding unit periphery The motion vector MV1 and motion vector MV2 of decoding region, export the position of the corresponding points C1 ' of C1, wherein A, and B is current decoding Point outside unit determines the terminating point of motion vector MV1 by the starting point of motion vector MV1 of A for A ', using B as motion vector The starting point of MV2 determines that the terminating point of motion vector MV2 is B ', and C1 is the representative locations of the 1st sub-block.

According to the similitude of triangle ABC2 and triangle A ' B ' C2 ', the periphery area decoder of current decoding unit is utilized The motion vector MV1 and motion vector MV2 in domain, export the position of the corresponding points C2 ' of C2, wherein A, and B is outside current decoding unit Point, determined by the starting point of motion vector MV1 of A the terminating point of motion vector MV1 for A ', using B as motion vector MV2 rise Initial point determines that the terminating point of motion vector MV2 is B ', and C2 is the representative locations of the 2nd sub-block.

According to the similitude of triangle ABC4 and triangle A ' B ' C4 ', the periphery area decoder of current decoding unit is utilized The motion vector MV1 and motion vector MV2 in domain, export the position of the corresponding points C1 ' of C4, wherein A, and B is outside current decoding unit Point, determined by the starting point of motion vector MV1 of A the terminating point of motion vector MV1 for A ', using B as motion vector MV2 rise Initial point determines that the terminating point of motion vector MV2 is B ', and C4 is the representative locations of the 4th sub-block.

Step 2: vector C1C1 ' is calculated by position C1 and position C1 ', vector C1C1 ' output is the 1st sub-block Motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX1.

Specifically, according to the coordinate (x of C1₁,y₁) and C1 ' coordinate (x₁’,y₁'), by the cross of following formula method export MVX1 To the longitudinal component MVX1y of component MVX1x and MVX1:

Vector C2C2 ' is calculated by position C2 and position C2 ', vector C2C2 ' output is the motion vector of the 2nd sub-block Predicted value or motion vector or motion vector predictor candidate or motion vector candidates MVX2.

Vector C4C4 ' is calculated by position C4 and position C4 ', vector C4C4 ' output is the motion vector of the 4th sub-block Predicted value or motion vector or motion vector predictor candidate or motion vector candidates MVX4.

Specifically, according to the coordinate (x of C4₄,y₄) and C4 ' coordinate (x₄’,y₄'), by the cross of following formula method export MVX4 To the longitudinal component MVX4y of component MVX4x and MVX4:

Embodiment 47

Following Module cycle twice, in the input of same MV1 and MV2, the different C points of the 1st sub-block and the 4th sub-block Under the input of position, the respective motion vector of the 1st sub-block and the 4th sub-block is obtained.

1	2
		3	4

Embodiment 48

Following Module cycle four times times, in the input of same MV1 and MV2, the input of the different C points position of all sub-blocks Under, obtain the respective motion vector of all sub-blocks.

1	2
		3	4

Embodiment 49

Specifically, the parallel degree of line segment AB and line segment A ' B ' are examined.The parallel degree of line segment AB and line segment A ' B ' are portrayed The degree of rotation of object, and the rotation between reference frame and present frame be more than to a certain degree after, the affine model of block grade is no longer It is applicable in.So the angle as line segment AB and line segment A ' B ' is less than given threshold value γ, is judged to examining rationally.Wherein the value of γ with The size of current decoding unit and change, the size of current decoding unit is bigger, and γ is smaller.

Embodiment 50

Specifically, the length ratio size of line segment AB and line segment A ' B ' are detected.Because of the length of line segment AB and line segment A ' B ' Scale features the zoom degree of object, and the scaling between reference frame and present frame be more than to a certain degree after, block grade Affine model is no longer applicable in.So when the length ratio size of line segment AB and line segment A ' B ' is in given range [1/ δ, δ], δ >=1 It is interior, it is judged to examining rationally, carries out step 2.Wherein the value of δ changes with the size of current decoding unit, and current decoding is single The size of member is bigger, and γ is smaller.

Embodiment 51

Specifically, the parallel degree of line segment AB and line segment A ' B ' are examined.The parallel degree of line segment AB and line segment A ' B ' are portrayed The degree of rotation of object, and after the rotation to a certain degree between reference frame and present frame, the affine model of block grade is no longer applicable in. So the angle as line segment AB and line segment A ' B ' is less than given threshold value γ, is judged to examining rationally.Wherein the value of γ is with current The size of sub-block and change, the size of current sub-block is bigger, and γ is smaller.

Embodiment 52

Specifically, the length ratio size of line segment AB and line segment A ' B ' are detected.Because of the length of line segment AB and line segment A ' B ' Scale features the zoom degree of object, and the scaling between reference frame and present frame be more than to a certain degree after, block grade Affine model is no longer applicable in.So when the length ratio size of line segment AB and line segment A ' B ' is in given range [1/ δ, δ], δ >=1 It is interior, it is judged to examining rationally, carries out step 2.Wherein the value of δ changes, the size of current sub-block with the size of current sub-block Bigger, γ is smaller.

Embodiment 52

A kind of motion vector guiding device that this example provides, as shown in Figure 9:

1	2
		3	4

Module S901, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, the 1st sub-block representative locations C1 point position coordinates；Its output includes: C1 ' point Position coordinates；

The function of the module are as follows: according to the similitude of triangle ABC1 and triangle A ' B ' C1 ', utilize current decoding unit The motion vector MV1 of decoding region and motion vector MV2, MV1 are not equal to MV2, export the position of the corresponding points C1 ' of C1 on periphery It sets, the terminating point of motion vector MV1 is determined by the starting point of motion vector MV1 of A for A ', using B as the starting of motion vector MV2 Point determines that the terminating point of motion vector MV2 is B '.

Module S902, input include: C1 ' point position coordinates；Its output includes motion vector MVX；

The function of the module are as follows: vector C1C1 ' is calculated by position C1 and position C1 ', vector C1C1 ' output is the 1st The motion vector predictor or motion vector or motion vector predictor candidate or motion vector of sub-block or current decoding unit are waited Select MVX.

Specifically, according to the coordinate of the coordinate (x, y) of C1 and C1 ' (x ', y '), by the transverse direction point of following formula method export MVX Measure the longitudinal component MVXy of MVXx and MVX:

Module S903, inputting includes: motion vector MV1, motion vector MV2, and the point position A outside current decoding unit is sat It marks, the B point position coordinates outside current decoding unit, the 2nd sub-block representative locations C2 point position coordinates；Its output includes: C2 ' point Position coordinates；

The function of the module are as follows: according to the similitude of triangle ABC and triangle A ' B ' C2 ', utilize current decoding unit The motion vector MV1 of decoding region and motion vector MV2, MV1 are not equal to MV2, export the position of the corresponding points C2 ' of C2 on periphery It sets, the terminating point of motion vector MV1 is determined by the starting point of motion vector MV1 of A for A ', using B as the starting of motion vector MV2 Point determines that the terminating point of motion vector MV2 is B '.

Module S904, input include: C2 ' point position coordinates；Its output includes motion vector MVX；

The function of the module are as follows: vector C2C2 ' is calculated by position C2 and position C2 ', vector C2C2 ' output is the 2nd The motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates MVX of sub-block.

Specifically, according to the coordinate of the coordinate (x, y) of C2 and C2 ' (x ', y '), by the transverse direction point of following formula method export MVX Measure the longitudinal component MVXy of MVXx and MVX:

Embodiment 53

The case where being bidirectional interframe predictive block for current decoding unit or current sub-block, the motion vector of either direction is led It out, can be before by the way of all embodiments.

Claims

1. a kind of motion vector deriving method, characterized by comprising:

According to the similitude of triangle ABC and triangle A ' B ' C ', the fortune of the periphery decoding region of current decoding unit is utilized Dynamic vector MV1 and motion vector MV2, MV1 export the position of the corresponding points C ' of C, wherein A not equal to MV2, and B is that current decoding is single Point outside member determines the terminating point of motion vector MV1 by the starting point of motion vector MV1 of A for A ', using B as motion vector MV2 Starting point determine that the terminating point of motion vector MV2 is B ', C is the generation of current decoding unit or current decoding unit intra sub-block Table position；

Vector CC ' is calculated by position C and position C ', vector CC ' output is in current decoding unit or current decoding unit The motion vector predictor or motion vector or motion vector predictor candidate or motion vector candidates of portion's sub-block.

2. motion vector deriving method as described in claim 1, it is further characterized in that including a reasonableness test step, institute Reasonableness test step is stated according to following criterion:

3. motion vector deriving method as described in claim 1, it is characterised in that including a kind of following situation:

From the left side of current decoding unit, decoding region obtains the motion vector MV1, from the top of current decoding unit Decoding region obtains the motion vector MV2；

From the left side of current decoding unit, decoding region obtains the motion vector MV2, from the top of current decoding unit Decoding region obtains the motion vector MV1.

4. motion vector deriving method as described in claim 1, it is characterised in that further include: the figure where current decoding unit The arrow of the movement described in obtaining is searched in decoding region on decoded reference picture on the periphery outside current decoding unit as in Measure MV1 and motion vector MV2.

5. motion vector deriving method as described in claim 1, it is further characterized in that: the current decoding unit intra sub-block It is at least two, the representative locations of each current decoding unit intra sub-block are respectively independent, inside each current decoding unit Sub-block is also respectively independent by the vector CC ' that position C and position C ' are calculated.

6. a kind of motion vector guiding device, it is characterised in that including,

Position export module: according to the similitude of triangle ABC and triangle A ' B ' C ', using current decoding unit periphery The motion vector MV1 and motion vector MV2, MV1 of decoding region export the position of the corresponding points C ' of C, wherein A, B not equal to MV2 Point outside for current decoding unit determines the terminating point of motion vector MV1 by the starting point of motion vector MV1 of A for A ', with B Determine that the terminating point of motion vector MV2 is B ' for the starting point of motion vector MV2, C is that current decoding unit or current decoding are single The representative locations of first intra sub-block；

Motion vector generation module: vector CC ' is calculated by position C and position C ', vector CC ' output is current decoding unit Or the motion vector predictor or motion vector or motion vector predictor candidate or movement arrow of current decoding unit intra sub-block Amount is candidate.

7. motion vector guiding device as claimed in claim 6, it is characterised in that including the position export module it Preceding addition reasonableness test module, the reasonableness test module is according to following criterion:

If meeting at least one above criterion, the vector of the position export module and motion vector generation module is determined The output of CC ' is effective.

8. motion vector guiding device as claimed in claim 6, it is characterised in that further include in the position export module Addition motion vector obtains module before, and it includes a kind of situation below that the motion vector, which obtains module:

9. motion vector guiding device as claimed in claim 6, it is characterised in that further include: in the position export module Motion-vector search export module is added before, and the motion-vector search export module is used for where current decoding unit The movement described in obtaining is searched in decoding region on decoded reference picture on the periphery outside current decoding unit in image Vector MV1 and motion vector MV2.

10. motion vector guiding device as claimed in claim 6, it is further characterized in that, son inside the current decoding unit Block is at least two, and the representative locations of each current decoding unit intra sub-block are respectively independent, in each current decoding unit Portion's sub-block is also respectively independent by the vector CC ' that position C and position C ' are calculated.