WO2012006889A1

WO2012006889A1 - Method and apparatus of temporal motion vector prediction

Info

Publication number: WO2012006889A1
Application number: PCT/CN2011/073167
Authority: WO
Inventors: Yu-Pao Tsai; Jian-Liang Lin; Yu-Wen Huang; Shaw-Min Lei
Original assignee: MediaTek Inc
Current assignee: MediaTek Inc
Priority date: 2010-07-12
Filing date: 2011-04-22
Publication date: 2012-01-19
Anticipated expiration: 2013-01-12
Also published as: EP2559252A1; US9124898B2; KR20130021388A; EP2559252A4; KR20150091271A; CN105704495A; US20150326876A1; CN102934434B; BR112012027263A2; US9961364B2; KR101709003B1; CA2794379A1; CA2794379C; US20120008688A1; JP2013529877A; BR112012027263B1; BR112012027263A8; CN105704495B; CN102934434A; AU2011278851B2

Abstract

An apparatus and method for motion vector prediction for a current block in a picture are disclosed. In video coding systems, the spatial and temporal redundancy is exploited using spatial and temporal prediction to reduce the information to be transmitted. Motion Vector Prediction (MVP) has been used to further conserve the bitrate associated with motion vector. In conventional temporal MVP, the predictor is often based on a single candidate such as the co-located motion vector in the previous frame/picture. If the co-located motion vector in the previous frame/picture does not exist, the predictor for the current block is not available. A techniquefor improved MVP is disclosed where the MVP utilized multiple candidates based on co-located motion vectors from future and/or past reference pictures. The candidates are arranged according to priority order to provide better availability of MVP and also to provide more accurate prediction. Furthermore, the MVP technique disclosed can be operated in a closed-loop fashion so that no additional side information or minimum additional side information is required.

Description

METHOD AND APPARATUS OF TEMPORAL MOTION

VECTOR PREDICTION

BACKGROUND OF THE INVENTION

Cross Reference To Related Applications [0001] The present invention claims priority to U.S. Provisional Patent Application,

No. 61/363,557, filed July 12, 2010, entitled "Video coding methods for B-frame referencing co-located motion vector" and U.S. Provisional Patent Application, No. 61/431,454, filed January 11, 2011, entitled "Improved advanced motion vector prediction".The U.S. Provisional Patent Applicationsarehereby incorporated by reference in its entirety.

Field of the Invention

[0002] The present invention relates to video coding. In particular, the present invention relates to coding techniques associated with motion vector prediction.

Description of the Related Art [0003] In video coding systems, spatial and temporal redundancy is exploited using spatial and temporal prediction to reduce the information to be transmitted. The spatial and temporal prediction utilizes decoded pixels from the same picture and reference pictures respectively to form prediction for current pixels to be coded. In a conventional coding system, side information associated with spatial and temporal prediction may have to be transmitted, which will take up some bandwidth of the compressed video data. The transmission of motion vectors for temporal prediction may require a noticeable portion of the compressed video data, particularly in low- bitrate applications. To further reduce the bitrate associated with motion vectors, a technique called Motion Vector Prediction (MVP) has been used in the field of video coding in recent years. The MVP technique exploits the statistic redundancy among neighboring motion vectors spatially and temporally.

[0004] When MVP is used, a predictor for the current motion vector is chosen and the motion vector residue is transmitted instead of the motion vector itself to conserve bitrate associated with motion vector transmission. The MVP scheme can be applied in a closed-loop arrangement where the predictor can be derived at the decoder based on decoded information and no side information has to be transmitted. Alternatively, side information can be transmitted explicitly in the bitstream to inform the decoder regarding the type of motion vector predictor selected. While MVP can be used for inter-coded blocks to conserve bandwidth, it can also be used for SKIP and DIRECT coded blocks to substantially reduce bitrate for the underlying blocks. In conventional temporal MVP, the predictor is often based on a single candidate such as the co- located motion vector in the previous frame/picture. If the co-located motion vector in the previous frame/picture does not exist, the predictor for the current block is not available. It is desirable to improve the performance of MVP so as to reduce the bitrate of the coding system. The improvement may be achieved by designing better MVP that can provide a more accurate prediction and improve the availability ofthe predictor. Furthermore, it is desirable that the MVP can be operated in a closed-loop fashion so that no side information or minimum side information is required.

BRIEF SUMMARY OF THE INVENTION

[0005] An apparatus and method for motion vector prediction for a current block in a picture are disclosed. In one embodiment according to the present invention, the apparatus and method for motion vector prediction comprise steps of receiving two or more motion vectors associated with at least one temporal block from a group consisting of one or more future reference pictures and one or more past reference pictures; determining a candidate set based on said two or more motion vectors, wherein the candidate set consists of at least two elements; arranging the candidate set in a priority order; and determining a motion vector predictor or a motion vector predictor candidate or a motion vector or a motion vector candidatefor the current block from the candidate set according to the priority order. Thepriority order is predefined in one embodiment according to the present invention and the priority order is determined according to an adaptive scheme in another embodiment according to the present invention. The step of determining the candidate set based onsaid two or more motion vectors may further comprises a step of deriving scaled motion vectors corresponding to the two or more motion vectors in an alternative embodimentaccording to the present invention. The candidate set may be selected from the two or more motion vectors, the scaled motion vectors, or a combination of both in various embodiments according to the present invention. The adaptive scheme may be based on a criterion selected from a group consisting of statistic of reconstructed motion vectors of previous blocks, partition type of the current block, correlation of the motion vectors, directions of motion vectors, distance of the motion vectors, and whether the motion vectors cross the current block in a case of temporal motion vectorin various embodiments according to the present invention. When the scaled motion vectors are used, the adaptive scheme can be based on whether directions of scaled motion vectorsare interpolation or extrapolation in an alternative embodiment according to the present invention.

[0006] In another embodiment according tothe present invention, the apparatus and method for motion vector prediction comprise steps of receiving two or more motion vectors associated with at least one temporal block from a first reference picture list; determining a candidate set consisting of at least two elements based on the two or more motion vectors; arranging the candidate set in a priority order; and determining a motion vector predictor or a motion vector predictor candidateor a motion vector or a motion vector candidate for the current block from the candidate set according to the priority order. Information associated with the priority order is incorporated in a sequence header, a picture header, or a slice header. The first reference picture list may be a list 0 reference picture or a list 1 reference picture. BRIEF DESCRIPTION OF DRAWINGS

[0007] Fig. 1 illustrates motion vector scaling of DIRECT mode prediction in B slice coding according to a prior art.

[0008] Fig. 2 illustrates motion vector scaling in B slice coding based on a co-located motion vectorof the first previous B pictureaccording to a prior art.

[0009] Fig. 3 illustrates an exemplary motion vector prediction using motion vectors of the co-located blocks in the list 0 and list 1 reference pictures with a pre-defined priority order.

[0010] Fig. 4 illustrates an example of pre-defined priority order for the four motion vectors of Fig. 3.

[0011] Fig. 5 illustrates an alternative pre-defined priority order for the four motion vectors of Fig. 3.

[0012] Fig. 6 illustrates an example of pre-defined priority order for the four motion vectors of Fig. 3, where only three of the four motion vectors are used as candidates for motion vector predictor.

[0013] Fig. 7 illustrates an alternative pre-defined priority order for the four motion vectors of Fig. 3, where only three of the four motion vectors are used as candidates for motion vector predictor.

[0014] Fig. 8 illustrates an example of pre-defined priority order for the four motion vectors of Fig. 3, where only two of the four motion vectors are used as candidates for motion vector predictor. [0015] Fig. 9 illustrates an alternative pre-defined priority order for the four motion vectors of Fig. 3, where only two of the four motion vectors are used as candidates for motion vector predictor.

[0016] Fig. 10A illustrates an example of motion vector prediction using scaled motion vectors derived for the current block with the current list equal tolist 0 and RefldxL0=0 reference picture.

[0017] Fig. 10B illustrates an example of motion vector prediction using scaled motion vectors derived for the current block with the current list equal to list 1 and RefldxLl=0 reference picture.

[0018] Fig. 11 illustrates an example of pre-defined priority order for the scaled four motion vectors of Fig. 10A.

[0019] Fig. 12 illustrates an example of pre-defined priority order for the scaled four motion vectors of Fig. 10B.

[0020] Fig. 13 illustrates an example of pre-defined priority order for scaled motion vectors wvLO_/ and wvLl_/ of Fig. 10A.

[0021] Fig. 14 illustrates an example of pre-defined priority order for scaled motion vectors wvLl and wvLO of Fig. 10A.

[0022] Fig. 15A illustrates an example of motion vector prediction using scaled motion vectorsderived for the current block with the current list equal tolist 0 and RefldxL0=0 reference picture and non-scaled motion vectors of co-located blocks from list 0 and list 1 reference pictures.

[0023] Fig. 15B illustrates an example of motion vector prediction using scaled motion vectorsderived for the current block with the current list equal tolist 1 and RefldxLl=0 reference picture and non-scaled motion vectors of the co-located block from list 0 and list 1 reference pictures.

[0024] Fig. 16 illustrates an example of pre-defined priority order for the scaled and non-scaled motion vectors of Fig. 15 A.

[0025] Fig. 17 illustrates an example of pre-defined priority order for the scaled and non-scaled motion vectors of Fig. 15B.

DETAILED DESCRIPTION OF THE INVENTION

[0026] In video coding systems, the spatial and temporal redundancy is exploited using spatial and temporal prediction to reduce the bitrate to be transmitted.The spatial prediction utilizes decoded pixels from the same picture to form prediction for current pixels to be coded. The spatial prediction is often operated on a block by block basis, such as 16x16 or 4x4 block for luminance signal in H.264/AVC Intra coding. In video sequences, neighboring pictures often bear great similarities, and simply using picturedifferences can effectively reduce the transmitted information associated with static background areas. Nevertheless, moving objects in the video sequence may result in substantialresidues and will require higher bitrate to code the residues. Motion Compensated Prediction (MCP) is a popular technique to exploit temporal correlation in video sequences.

[0027] Motion compensated prediction can be used in a forward prediction fashion, where a current picture block is predicted using a decoded picture or pictures that are prior to the current picture in the display order. In addition to forward prediction, backwardprediction can also be used to improve the performance of motion compensated prediction. The backwardprediction utilizes a decoded picture or pictures afterthe current picture in the display order.Sincethe first version of H.264/AVC was finalized in 2003, forward prediction and backward prediction have been extended to list 0 prediction and list 1 prediction, respectively, where both list 0 and list 1 can contain multiple reference pictures prior to or past the current picture in the display order. The following describes the default reference picture list reconstruction.For list 0, reference pictures prior to the current picture have lower reference picture indices than those past the current picture. For list 1, reference pictures past the current picture have lower reference picture indices than those prior to the current picture. For both list 0 and list 1, after applying the previously described rule, the temporal distance is also considered in determining the reference picture index. Generally speaking, a reference picture closer to the current picture has a lower reference picture index. For example, assume the current picture is picture 5, andpictures 0, 2, 4, 6, and 8 are reference pictures, where the numbers denote the display order. The list 0 reference pictures with ascending reference picture indices and starting with index equal to zero are 4, 2, 0, 6, and 8. The list 1 reference pictures with ascending reference picture indices and starting with index equal to zero are 6, 8, 4, 2, and 0. The reference picture with index equal to 0 is called co-located picture, and in this example with picture 5 as the current picture, picture 6 is the list 1 co- located picture, and picture 4 is the list 0 co-located picture. When a block in a list 0 or list 1 co-located picture has the same block location as the current block in the current picture, it is called a list 0 or list 1 co-located block, or called a co-located block in list 0 or list l .The unit used for motion estimation mode in earlier video standards such as MPEG-1, MPEG-2 and MPEG-4 is primarily based on macroblock. For H.264/AVC, the 16x16 macroblock can be segmented into 16x16, 16x8, 8x16 and 8x8 blocks for motion estimation. Furthermore, the 8x8 block can be segmented into 8x8, 8x4, 4x8 and 4x4 blocks for motion estimation. For the High Efficiency Video Coding (HEVC) standard under development, the unit for motion estimation/compensation mode is called Prediction Unit (PU), where the PU is hierarchically partitioned from a maximum block size. The MCP type is selected for each slice in the H.264/AVC standard. A slicethat the motion compensated prediction is restricted to the list Oprediction is called a P-slice. For a B-slice, the motion compensated prediction also includes the list 1 prediction and the bidirectional predictionin addition to the list Oprediction.

[0028] In video coding systems, the motion vector and coded residues are transmitted to a decoder for reconstructing the video at the decoder side. Furthermore, in a system with flexible reference picture structure, the information associated with the selected reference pictures may also have to be transmitted. The transmission of motion vectors may require a noticeable portion of the transmitted bitrate, particularly in low-bitrate applications or in systems where motion vectors are associated with smaller blocks or high motion accuracy. To further reduce the bitrate associated with motion vector, a technique called Motion Vector Prediction (MVP) has been used in the field of video coding in recent years. The MVP technique exploits the statistic redundancy among neighboring motion vectorsspatially and temporally. When MVP is used, a predictor for the current motion vector is chosen and the motion vector residue, i.e., the difference between the motion vector and the predictor, is transmitted. The MVP scheme can be applied in a closed-loop arrangement where the predictor is derived at the decoder based on decoded information and no side information has to be transmitted. Alternatively, side information can be transmitted explicitly in the bitstream to inform the decoder regarding the type of motion vector predictor selected.

[0029] In the H.264/AVC standard,there is also aSKIP mode in additional tothe conventional Intra and Inter modes for macroblocks in a Pslice. The SKIP mode is a very effective method to achieve large compression since there is no quantized error signal, no motion vector, nor reference index parameter to be transmitted. The only information required for the 16x16 macroblock in the SKIP mode is a signal to indicate the SKIP mode being used and therefore substantial bitrate reduction is achieved. The motion vector used for reconstructing the SKIP macroblock is similar to the motion vector predictor for a macroblock.In the H.264/AVC standard, four different types of inter-prediction are supported for B slices includinglist 0, list 1, bi- predictive, and DIRECT prediction, where list 0 and list 1 refer to prediction using reference picture group 0 and group 1 respectively. For the bi-predictive mode, the prediction signal is formed by a weighted average of motion-compensated list 0 and list 1 prediction signals. The DIRECT prediction mode is inferred from previously transmitted syntax elements and can be either list 0 or list 1 prediction or bi- predictive. Therefore, there is no need to transmit information for motion vector in the DIRECT mode. In the case that no quantized error signal is transmitted, the DIRECT macroblock mode is referred to as B SKIP mode and the block can be efficiently coded.

[0030] In HEVC being developed, some improvement of motion vector prediction over the H.264/AVC is being considered. In this disclosure, a system and method of motion vector prediction for Bframe/picture/slice based on temporal blocks in the past and/or the future reference pictures are disclosed.The motion vector for a current block is predicted by the motion vectors oftemporal blocks in the past and/or the future reference pictures effectively so thatthe coding efficiency for the motion vector can be improved. The temporal motion vectors are considered as candidates of predictor for the current block and the candidates are arranged in priority order. The candidate with higher priority order will be considered as predictor ahead of a candidate with a lower priority order. The advantage of priority based MVP derivation is to increase the chance that the temporal MVP candidate for a current block exists. For example, the conventional video coding system only considers the co-located motion vector in the previous frame/picture as the candidate. If that candidate does not exist, then the coding system will consider the MVP candidate for the co-located block to be unavailable. Therefore, it is desirable to increase the availability of MVP candidate so as to improve the coding efficiency of the coding system.

[0031] In the H.264/AVC standard, the temporal DIRECT mode is used for B slices where the motion vectors for a current block 110 in the B slice is derived from the motion vector of the co-located block 120 in the first list 1 reference picture as shown in Fig. 1. The motion vector derivation for the temporal DIRECT mode is described in "Direct Mode Coding for Bipredictive Slices in the H.264 Standard", authored by Tourapis et al., in IEEE Trans, on Circuits and Systems for Video Technology, Vol. 15, No. 1, pp.119-126, Jan. 2005. The motion vector for the co-located block of the first List 1 reference is denoted as MV . The motion vectors for the current block are denoted as MVLO and MVn with respect to the list 0 reference picture and list 1 reference picture. The temporal distance between the current picture and the list 0 reference picture is denoted as TD_B and the temporal distance between the list 0 reference picture and the list 1 reference picture is denoted as TD_D. The motion vectors for the current block can be derived according to:

M o (1 )

MV (2) The above equations were later replaced by:

(16384+a&s(7¾/2)

Λ — j )

TD_D

ScaleFactor = dip(-lG24,1023, (TD_B x X + 32) » 6) (4) ¾o = (ScaleFactor x W + 128) » 8 (5)

MV = A¾ - MV (6) so that X and ScaleFactor can be pre-computed at the slice/picture level. In the temporal DIRECT mode, the motion vector prediction is only based on the motion vector for the co-located block of the first list 1 reference.

[0032] In another prior art, entitled "RD Optimized Coding for Motion Vector Predictor Selection", by Laroche et ah, in IEEE Trans, on Circuits and Systems for Video Technology, Vol. 18, No. 12, pp.1681-1691, Dec. 2008, motion vector prediction selection based on motion vector competition is disclosed. The motion vector competition scheme uses RD optimization to determine the best motion vector predictor from motion vector predictor candidates. For example, as shown in Fig. 2, the temporal motion vector predictor candidates may include the list 0 motion vector corresponding to the co-located block in the list 1 co-located picture Refi, and the list 0 and list 1 motion vectors for a co-located block in the list 0 co-located picture, B-l. The list 0 motion vectors corresponding to the co-located block in the list 1 co-located picture Refi can be calculated in the same way as defined in the H.264/AVC standard: m^vcoh

mvi = — x d_LQ (7)

[0033] The list 0 and list 1 motion vectors for a co-located block in the list 0 co- located picture, B-\ , can be used to derive motion vector predictor for the current block. If only the co-located motion vector ^{nu L0}^-n.o in picture B-l pointing to a forward P-picture exists, the motion predictors tflV^ ⁰ and tflV^¹ can be calculated according to:

m_vLl ₌ - LO _{χ (d} _ _{d (m}

[0034] The motion vector™ ^c0^ ^"'ω is depicted in Fig. 2 and dwB - i is the temporal distance between the forward P-frame and frame B-\ . In the case of backward prediction, the predictors tflV^⁰ and JflV^¹ can be calculated according to:

f₁ ^{, v}col_B→

m_v l _{= χ} y _ ^ j _{( 1 2)}

[0035] The motion vector ^'-B- S-H _S the co-located motion vector in picture B-\ pointing to the past P-frame as depicted in Fig. 2. Depending on the availability of motion vectors ~ and ' "- 'it, the corresponding predictors m equations (7)-(12) can be used for the current block and the RD optimization is applied to select the best motion vector predictor. The motion vector prediction scheme according to Laroche et al. will require side information to be transmitted to the decoder side to indicate the particular motion vector predictor selected. The transmission of side information associated with the selected motion vector predictor will consume some bandwidth. Regardless whether the motion vector competition scheme is enabled or disabled, temporal motion vector prediction can be beneficial for reducing motion vector residues. It is desirable to develop a temporal motion vector prediction technique to enhance any temporal motion vector predictor when the motion vector competition is not used, as well as to enhance any temporal motion vector predictor candidate when the motion vector competition is used. Moreover, it is also desirable that the enhanced temporal motion vector prediction has no additional side information or minimum additional side information.

[0036] Accordingly, a motion vector prediction technique for a current block in a picture is developed.The motion vector prediction technique of the present invention can be used to determine a motion vector predictor, a motion vector predictor candidate, a motion vector, or a motion vector candidate for the current block. The technique uses motion vectors associated with temporal blocks from the future and past reference pictures as candidates forpredictor and selects a predictor according to a priority order. Fig. 3 illustrates exemplary motion vector prediction using motion vectors of co-located blocks in the pastand the future reference pictures withaccording to apriority order. The example of future picture used in Fig.3 is the picture belongmg to list lwith reference picture index RefldxLl equal to 0 and the pictureis labeled as "/". The example of past picture used in Fig. 3is the picture belonging to list 0 withreference picture index RefldxLO equal to 0 andthe pictureis labeled as "j ".Furthermore, the current pictureis labeled as "k ", the picturebelongingto list 0 with RefldeL0>\ is labeled as "z ", and the picture belonging to list 1 with RefldeLl>\ is labeled as "w".The motion vectors mvLO and mvLl for a current block 310 are determined fromwotion vectors mvLO_j, mvLl_j, mvLOi, andwvLl/ of co-located block 320 of picture/and co-located block 330 of picture/respectively according to apriority order. The motion vectors mvLO_j and wvLl, are referring to motion vectors for co- located block 320 of picture/ pointing to a picturein list 0 and a picturein list 1 respectively. The motion vectors mvLOj and mvLlj are referring to motion vectors for co-located block 330 of picture/ pointing to a picturein list 0 and a picturein list 1 respectively. These motion vector candidates will be arranged in apriority order so that a predictor can be chosen accordingly.

[0037] Fig. 4 illustrates an example of pre-defined priority order for the four motion vector candidates of Fig. 3. The determination of motion vector predictorbased on co- located motion vectors in the future and past reference pictures for the current block 310to use a reference picture in list 0 is shown as follows: • If mvLOi exists and the corresponding reference picture (pic. id = i-1) is in the list 0 of the current picture,

then wvLO = wvLO_/ (referencepic. id = i - X) ;

• Else if mvLl_j exists and the corresponding reference picture (pic. id = m) is in the list 0 of the current picture,

then wvLO = mvLl_j (referencepic. id = m) ;

• Else if mvLO_j exists and the corresponding reference picture (pic. id =j) is in the list 0 of the current picture,

then wvLO = wvLO/ (referencepic. id=j ) ;

• Else if mvLli exists and the corresponding reference picture (pic. id = I) is in the list 0 of the current picture,

then wvLO = mvLlj (reference pic. id = I ) ;

• Else,

mvLO is not available.

The pre-defined priority order shown above illustrates an example of determining motion vector predictoror motion vector predictor candidate wvLOfor the current block 310 to use a reference picture in list 0 and the candidates for the predictor are based on co-located motion vectors in list 0 and list 1 reference pictures. A skilled person in the art may use other pre-defined priority orders to achieve the same or similar goal.

[0038] Fig. 5 illustrates another example of pre-defined priority order for the four motion vector candidates of Fig. 3. The determination of motion vector predictor based on co-located motion vectors in the future and past reference pictures for the current block 310 to use a reference picture in list 1 is shown as follows:

• If mvLl_j exists and the corresponding reference picture (pic. id = m) is in the list lof the current picture,

then mvL\ = mvLl_j (referencepic. id = m);

• Else if mvLOi exists and the corresponding reference picture (pic. id = i- l)is in the list lof the current picture,

then mvL\ = mvLOj (referencepic. id = i - 1);

• Else if mvLlj exists and thecorresponding reference picture (pic. id = l)is in the list lof the current picture,

then mvL\ = mvLli (reference pic. id = I);

• Else if mvLO_j exists and the corresponding reference picture (pic. id =j)is in the list lof the current picture,

then mvL\ = mvLO_j (referencepic. id=j);

• Else,

mvLl is not available.

The pre-defined priority order shown above illustrates an example of determining motion vector predictor or motion vector predictor candidate wvLlfor the current block 310 to use a reference picture in list 1 and the candidates for the predictor are based on co-located motion vectors in list 0 and list 1 reference pictures. A skilled person in the art may select other pre-defined priority orders to achieve the same or similar goal.

[0039] While all four vectors are considered as candidates for the predictor or predictor candidate of the current block, not all motion vector candidates have to be used. For example, three of the four motion vectors in Fig. 3 can be used to determine the motion vector predictoror motion vector predictor candidate for the current block 310 to use a reference picture inlist 0 as shown in Fig. 6:

• If mvLOj exists and the corresponding reference picture is in the list 0 of the current picture,

then wvLO = wvLO_/ (referencepic. id= i - 1);

• Else if mvLl_j exists and the corresponding reference picture is in the list 0 of the current picture,

then wvLO = mvLl_j (referencepic. id= m);

• Else if mvLO_j exists and the corresponding reference picture is in the list 0 of the current picture,

then mvLO = mvLO_j (referencepic. id=j);

• Else,

mvLO is not available.

[0040] Similarly, three of the four motion vectors in Fig. 3 can be used to determine the motion vector predictor or motion vector predictor candidate for the current block 310 to use a reference picture inlist 1 as shown in Fig. 7:

• If mvL\_j exists and the corresponding reference picture is in the list lof the current picture,

then mvLl = mvLl_j (referencepic. id= m);

• Else if mvLOi exists and the corresponding reference picture is in the list lof the current picture, then mvL\ = wvLO_/ (referencepic. id = i - 1);

• Else if mvLlj exists and the corresponding reference picture is in the list lof the current picture,

then wvLl = mvLlj (reference pic. id = I );

• Else,

mvLl is not available.

[0041] In another example, two of the four motion vectors in Fig. 3 can be used to determine the motion vector predictor or motion vector predictor candidate for the current block 310 to use a reference picture in list 0 as shown in Fig. 8:

• If mvLOi exists and thecorresponding reference picture is in the list 0 of the current picture,

then wvLO = wvLO_/ (referencepic. id = i - 1);

• Else if mvLl_j exists and thecorresponding reference picture is in the list 0 of the current picture,

then wvLO = mvLl_j (referencepic. id = m);

• Else,

mvLO is not available.

[0042] Similarly, two of the four motion vectors in Fig. 3 can be used to determine the motion vector predictor or motion vector predictor candidate for the current block 310 to use a reference picture in list 1 as shown in Fig. 9:

• If mvLl_j exists and the corresponding reference picture is in the list lof the current picture,

then wvLl = mvLl_j (referencepic. id = m);

• Else if mvLOi exists and the corresponding reference picture is in the list lof the current picture,

then mvLl = mvLOj (referencepic. id = i - 1);

• Else,

mvLl is not available.

[0043] It is not restricted to use at least one motion vector associated with a temporal block from a list 0 reference picture and at least one motion vector associated with a temporal block from a list 1 reference picture to determine the motion vector predictor or motion vector predictor candidate for the current block. In some other embodiments, a priority order is incorporated in the bitstream, such as in a sequence header, a picture header, or a slice header; and the motion vector predictor or motion vector predictor candidate for the current block 310 is determined from two motion vectors wvLO_/ and mvLlj associated with the list 1 co-located block 330 according to the priority order. In another example, the motion vector predictor or motion vector predictor candidate for the current block 310 is determined from two motion vectors mvLO_j and mvLl_j associated with the list 0 co-located block 320 according to the priority order incorporated in a sequence header, picture header, or slice header.

[0044] While the examples illustrated above use co-located motion vectors of reference pictures in the future with RefldxLl = 0 and in the past with RefldxLO = 0 as candidates of motion vector predictor or motion vector predictor candidate for the current block, scaled co-located motion vectors of the corresponding reference pictures may also be used as candidates. A method of motion vector scaling has been described in Figs. 1 and 2 and associated text. The same scaling method can be used to derive the scaled motion vectors for the present invention. Fig. 10A illustrates an example ofmotion vector prediction based on scaled motion vectors of list 0 and list 1 co-located blockswhen the current list islist 0 and RefldxL0=0. The scaled motion vectors corresponding to wvLO/, mvLl_j, mvLO_j, and wvLl/ are labeled as 1002, 1004, 1006 and 1008 respectively. The scaled motion vector represents the motion vector between the current block and thereference picture in list 0 with RefldxL0=0 by scaling the list 0 or list 1 co-located motion vector. Therefore, instead of using mvLOj, mvLl_j, mvLO_j, and mvLli as the candidates for motion vector predictor or motion vector predictor candidate, the scaled motion vectors 1002, 1004, 1006 and 1008 are used as the candidates for motion vector predictor or motion vector predictor candidate. The scaled motion vectors may provide better prediction than the non- scaled motion vectors. Fig. 10B illustrates an example ofmotion vector prediction based on scaled motion vectors of list 0 and list 1 co-located blockswhen the current list islist land RefldxLl=0. The scaled motion vectors corresponding to mvLl_j, mvLOi, mvLlj and wvLO, are labeled as 1012, 1014, 1016 and 1018 respectively. The scaled motion vector represents the motion vector between the current block and thereference picture in list 1 with RefldxLl=0 by scaling the list 0 or list 1 co-located motion.Therefore, instead of using wvLl ,wvL0/, mvLli and mvLO_j as the candidates for motion vector predictor or motion vector predictor candidate, the scaled motion vectors 1012, 1014, 1016 and 1018 are used as the candidates for motion vector predictor or motion vector predictor candidate. The scaled motion vectors may provide better prediction than the non-scaled motion vectors. [0045] Similar to the case for non-scaled motion vectors, a pre-defined priority order can be assigned to the scaled motion vectors for list 0 motion vector prediction. The list 0 reference picture index RefldxLO can be implicitly derived or explicitly transmitted in the bitstream. In the case that the current list is list 0 and RefldxL0=0, the pre-defined priority order for the scaled motion vectors of Fig. 10A is shown in Fig. 11. The motion vector predictor or motion vector predictor candidate is determined according to the following priority order:

• If mvLOi exists,

then mvLO = scaled wvLO_/ (motion vector 1002);

• Else if mvLl_j exists,

then mvLO = scaled mvLl_j (motion vector 1004);

• Else if mvLO_j exists,

then mvLO = scaled wvLO, (motion vector 1006);

• Else if mvLl /exists,

then mvLO = scaled mvLli (motion vector 1008);

• Else,

mvLO is not available.

[0046] In another embodiment, the list 0 motion vector prediction (the current list is list 0 and RefIdxL0=0) for driving the motion vector predictor or motion vector predictor candidate for the current block 310 is determined according to the following priority order:

• If wvLO/ exists,

then mvLO = scaled mvLOi (motion vector 1002);

• Else if mvLli exists,

then mvLO = scaled mvLlj (motion vector 1008);

• Else,

mvLO is not available.

In this case, only the scaled motion vectors associated with the list 1 co-located block 330 areconsidered. Information associated with the priority order is incorporatedin a sequence header, a picture header, or a slice header.

[0047] Yet another embodiment of the list 0 motion vector prediction (the current list is list 0 and RefldxL0=0) determines the motion vector predictor or motion vector predictor candidate for the current block 310 according to the following priority order:

• If mvLl_j exists, then mvLO = scaled mvLl_j (motion vector 1004);

• Else if mvLO_j exists,

then mvLO = scaled wvLO, (motion vector 1006);

• Else,

wvLO is not available.

In this case, only the scaled motion vectors associated with the list 0 co-located block 320 are considered. Information associated with the priority order is incorporatedin a sequence header, a picture header, or a slice header.

[0048] A pre-defined priority order can also be assigned to the scaled motion vectors for list 1 motion vector prediction.The List 1 reference picture index RefldxLl can be implicitly derived or explicitly transmitted in the bitstream. In the case that the current list is list 1 and RefldxLl=0, the pre-defined priority order for the scaled motion vectors of Fig. 10B is shown in Fig. 12. The motion vector predictor or motion vector predictor candidate is determined according to the following priority order:

• If mvLl_j exists,

then mvLl = scaled mvLl_j ;

• Else if mvLOi exists,

then mvLl = scaled wvLO/ ;

• Else if mvLli exists,

then mvLl = scaled mvLl ,

• Else if mvLO_j exists,

then mvLl = scaled mvLO_j ;

• Else,

mvLl is not available.

Similarly, the list 1 motion vector prediction (the current list is list 1 and RefldxLl=0) may only consider scaled motion vectors associated with the list 0 co-located block 320 (i.e. motion vectors 1012 and 1018) according to a pre-defined priority order incorporated in the bitstream; or it may only consider scaled motion vectors associated with the list 1 co-located block 330 (i.e. motion vectors 1014 and 1016) according to a pre-defined priority order incorporated in the bitstream.

[0049] It is also possible to choose two or three scaled motion vectors in Fig. 11 or Fig. 12 as candidates for motion vector predictor or motion vector predictor candidate, and the motion vector predictor or motion vector predictor candidate for the current block is determined from the candidates according to a priority order. For example, the candidates candidate include two motion vectors 1002 and 1004 of Fig. 11, and according to a priority order, the motion vector 1002 will be considered before the motion vector 1004. In another embodiment, the candidates include three motion vectors 1012, 1014, and 1016 of Fig. 12, and according to a priority order, the motion vector 1012 will be first considered, then the motion vector 1014, and last, the motion vector 1016.

[0050] As previously described, while the example in Fig. 11 considers all scaled motion vectors corresponding to list 0 and list 1 co-located motion vectors as candidates for motion vector predictor or motion vector predictor candidate, the candidates can be limited to scaled motion vectors corresponding to the co-located block in list 0 or list 1 only.As before, the list 0 reference picture index RefldxLO can be implicitly derived or explicitly transmitted in the bitstream. In the case that the current list is list 0 and RefldxL0=0, only the scaled motion vectors associated with the co-located block in list lmay be considered ascandidates. An example of predefined priority order for this case is shown in Fig. 13. Information of the current list is list 0 can be incorporated or defined in each prediction unit (PU). The motion vector predictor or motion vector predictor candidate is determined according to the following priority order:

• If wvLO/ exists,

then wvLO = scaled wvLO_/ ;

• Else if mvLli exists,

then mvLO = scaled mvLl ,

• Else,

mvLO is not available.

[0051] In the above example, if only the motion vectors associated with the co-located block in list Oare considered ascandidates, an example of pre-defined priority order for this case is shown in Fig. 14. The motion vector predictor or motion vector predictor candidate is determined according to the following priority order:

• If mvLl_j exists,

then mvLO = scaled mvLl_j ;

• Else if mvLO_j exists,

then mvLO = scaled mvLO_j ;

• Else,

mvLO is not available. [0052] In the above examples, the motion vector prediction is either based on scaled motion vectors or non-scaled motion vectors of list 0 and list 1 co-located motion vectors. The scaled motion vectors and the non-scaled motion vectors of list 0 and list 1 co-located motion vectors can be combined to provide more choices for better motion vector prediction. Fig. 15A illustrates an example ofmotion vector prediction using scaled motion vectors and non-scaled motion vectors. Both scaled motion vectors 1002, 1004, 1006 and 1008 between the current block 310 and the reference picture in list 0 with picture id=j, and non-scaled motion vectors wvLO/, wvLl , wvLO , and mvLli of the co-located blocks in the future and past reference pictures are used as the candidates for motion vector predictor or motion vector predictor candidate. Fig. 15B illustrates an example ofmotion vector prediction using scaled motion vectors and non-scaled motion vectors. Both scaled motion vectors 1012, 1014, 1016 and 1018 between the current block 310 and the reference picture in list 1 with picture id=l, and non-scaled motion vectorswvLl , wvLO/, , mvLlj and wvLO are used as the candidates for motion vector predictor or motion vector predictor candidate.

[0053] Fig. 16 illustrates an example of motion vector prediction using both scaled motion vectors and non-scaled motion vectors of the co-located block in the future and past reference pictures as the candidates for motion vector predictor as shown in Fig. 15 A. The candidates for motion vector predictor or motion vector predictor candidate include scaled motion vectors 1002, 1004, 1006 and 1008 between the current block 310 and the reference picture inlist 0 with picture id=j, and non-scaled motion vectors wvLO/, wvLl , wvLO , and wvLl/ of co-located blocks in the future and past reference pictures are used as the candidates for motion vector predictor or motion vector predictor candidate. The determination of the predictor or predictor candidate is according to the following pre-defined priority order:

• If wvLO/ exists and the corresponding reference picture (pic. id = i - 1) is in the list 0 of the current picture,

then wvLO = wvLO/ (referencepic. id= i - 1);

Else,

wvLO = scaledwvLO/ (referencepic. id=j);

• Else if mvL\_j exists and the corresponding reference picture (pic. id = w+i)is in the list 0 of the current picture,

then mvLO = mvLl_j (referencepic. id= m+\);

Else, mvLO = scaledwvLl (referencepic. id =j);

• Else if mvLO_j exists and the corresponding reference picture (pic. id =j)is in the list 0 of the current picture,

then mvLO = mvLO_j (referencepic. id =j);

Else,

wvLO = scaledwvLO (referencepic. id =j);

• Else if mvLli exists and the reference picture (pic. id = I) is in the list 0 of the current picture,

then wvLO = mvLXi (reference pic. id = I);

Else,

wvLO = scaled wvLl _/(referencepic. id =j);

• Else,

mvLO is not available.

[0054] Fig. 17 illustrates an example of motion vector prediction using both scaled motion vectors and non-scaled motion vectors of the co-located block in the future and past reference pictures as the candidates for motion vector predictoras shown in Fig. 15B. The candidates for motion vector predictor or motion vector predictor candidate include scaled motion vectors 1012, 1014, 1016 and 1018 between the current block 310 and the reference picture in list lwith picture id=l, and non-scaled motion vectors mvLOi, mvLX_j, mvLO_j, and mvLli of the co-located block in the future and past reference pictures are used as the candidates for motion vector predictor or motion vector predictor candidate. The determination of the predictor or predictor candidate is according to the following pre-defined priority order:

• If mvLX_j exists and the corresponding reference picture (id = m+X)is in the list lof the current picture,

then mvLX = mvLX_j (referencepic. id = m+X);

Else,

mvLX = scaled mvLX_j (referencepic. id = I);

• Else if mvLOi exists and the corresponding reference picturefz^'t/ = i - 1) is in the list lof the current picture,

then mvLX = mvLOj (referencepic. id = i - 1);

Else,

mvLX = scaledwvLO_/ (referencepic. id = I);

• Else if mvLXi exists and the corresponding reference picture(pic. id = I) is in the list lof the current picture,

then mvL\ = wvLl_/ (reference pic. id = I);

Else,

mvLl = scaledwvLl _/(referencepic. id = I);

• Else if mvLO_j exists and thecorresponding reference picture (pic. id =j)is in the list lof the current picture,

then wvLl = mvLO_j (referencepic. id =j);

Else,

wvLl = scaledwvLO/ (referencepic. id = I);

• Else,

wvLl is not available.

[0055] In the above examples of motion vector prediction according to a pre-defined priority order, a respective priority orderis used in each example to illustrate the process of determining a motion vector predictor or a motion vector predictor candidatefrom the motion vector candidates. The particularpriority order used is by no means construed as a limitation to the present invention. A skilled person in the field may choose different priority orders for the motion vector candidates to implement the present invention. Furthermore, while the above examples illustrate thatthe order of motion vector among the candidates is determined according to a predefined priority order, the priority orderof the candidates can also be performedaccording to an adaptive scheme.The scheme of adaptive priority ordering can be based on the statistic of the reconstructed motion vectors of previous blocks, the partition type of current block, the correlation of the motion vectors, the directions of motion vector, whether the direction of scaled motion vector is interpolation or extrapolation, the distance of the motion vectors, or whether themotion vector crosses current block in the case of temporal motion vector. Also, the adaptive scheme may also be based on a combination of two or more of the factors mentioned above.

[0056] When the statistic of the reconstructed motion vectors of previous blocks is used for the adaptive scheme, the statistic may be associated with the counts of the motion vector candidates as an example. The priority order is adapted to the counts of the motion vector candidates, where the motion vector candidate having a higher count will be assigned a higher priority for motion vector predictor. When the partition type of current block is used for the adaptive scheme,for example, ifacurrent coding unit of size 2Nx2N is divided into two rectangular prediction units of size Nx2N and the current block is the left prediction unit, the motion vector with higher similarity to the left neighbor of the current coding unit will be assigned a higher priority; if a current coding unit of size 2Nx2N is divided into two rectangular prediction units of size Nx2N and the current block is the right prediction unit, the motion vector with higher similarity to the above right neighbor of the current coding unit will be assigned a higher priority. When the correlation of the motion vectors is used for the adaptive scheme, the motion vector with higher correlation will be assigned with a higher priority. For example, if two motion vectors in the priority list are exactly the same, the motion vector is considered to have higher correlation. When the direction of motion vector is used for the adaptive scheme, the motion vector pointing to the direction of the target reference picture, as an example, will be assigned with a higher priority. When the direction of scaled MV being interpolation or extrapolation is used for the adaptive scheme, the scaled motion vectors with the process of interpolation, as an example, will be assigned with a higher priority. When the distance of the motion vectors is used for the adaptive scheme, a shorter temporal distance for the motion vector from a current block to the target reference picture, as an example, will be assigned with a higher priority. When the coverage of a temporal motion vector is used as the adaptive scheme, the coverage of the motion vector over a current block will be assigned a higher priority.

[0057] It is noted that the present invention can be applied to not only Inter mode but also Skip, Direct, and Merge modes. In the Inter mode, given a current list, a motion vector predictor is used to predict the motion vector of a PU, and a motion vector residue is transmitted. The current invention can be applied for deriving the motion vector predictor when the motion vector competition scheme is not used or for deriving the motion vector predictor candidate when the motion vector scheme is used. As for the Skip, Direct, and Merge, they can be regarded as special cases of the Inter mode where the motion vector residue is not transmitted and always inferred as zero. In these cases, the current invention can be applied for deriving the motion vector when the motion vector competition scheme is not used or for deriving the motion vector candidate when the motion vector scheme is not used.

[0058] Embodiment of motion vector prediction according to the present invention as described above may be implemented in various hardware, software codes, or a combination of both. For example, an embodiment of the present invention can be a circuit integrated into a video compression chip or program codes integrated into video compression software to perform the processing described herein. An embodiment of the present invention may also be program codes to be executed on a Digital Signal Processor (DSP) to perform the processing described herein. The invention may also involve a number of functions to be performed by a computer processor, a digital signal processor, a microprocessor, or field programmable gate array (FPGA). These processors can be configured to perform particular tasks according to the invention, by executing machine-readable software code or firmware code that defines the particular methods embodied by the invention. The software code or firmware codes may be developed in different programming languages and different format or style. The software code may also be compiled for different target platform. However, different code formats, styles and languages of software codes and other means of configuring code to perform the tasks in accordance with the invention will not depart from the spirit and scope of the invention.

[0059] The invention may be embodied in other specific forms without departing from its spirit or essential characteristics. The described examples are to be considered in all respects only as illustrative and not restrictive. The scope of the invention is, therefore, indicated by the appended claims rather than by the foregoing description. All changes which come within the meaning and range of equivalency of the claims are to be embraced within their scope.

Claims

1. A method of deriving a motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for a current block in a picture, the method comprising:

receiving two or more motion vectors associated with at least one temporal block from a group consisting of one or more list 0 reference pictures and one or more list 1 reference pictures;

determining a candidate set based onsaid two or more motion vectors, wherein the candidate set consists of at least two elements;

arranging the candidate set in a priority order; and

determiningthe motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for the current block from the candidate set according to the priority order.

2. The method of Claim 1, wherein the priority order is a pre-defined priority order.

3. The method of Claim 2, wherein information associated with the pre-defined priority order is incorporated in a sequence header, a picture header, or a slice header.

4. The method of Claim 1, wherein the priority order is determined according to an adaptive scheme.

5. The method of Claim 4, wherein the adaptive scheme is based on a criterion selected from a group consisting of:

statistic of reconstructed motion vectors of previous blocks,

partition type of the current block,

correlation of the motion vectors,

directions of motion vectors,

distance of the motion vectors, and

whetherthemotion vectors crossthe current block in a case of temporal motion vector.

6. The method of Claim 1 , wherein the candidate set is selected from said two or more motion vectors.

7. The method of Claim 1 , wherein saiddetermining the candidate set comprises deriving scaled motion vectors corresponding to said two or more motion vectors, wherein the candidate set is selected from the scaled motion vectors or a combination of said two or more motion vectors and the scaled motion vectors.

8. The method of Claim 7, wherein the candidate set is selected from the scaled motion vectors.

9. The method of Claim 7, wherein the candidate set is selected from said two or more motion vectors and the scaled motion vectors.

10. The method of Claim 7,wherein the priority order is according to an adaptive scheme based on whether directions of scaled motion vectorsare interpolation or extrapolation.

1 1. An apparatus of deriving a motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for a current block in a picture, the apparatus comprising:

means for receiving two or more motion vectors associated with at least one temporal block from a group consisting of one or more list 0 reference pictures and one or more list 1 reference pictures;

means for determining a candidate set based onsaid two or more motion vectors, wherein the candidate set consists of at least two elements;

means for arranging the candidate set in a priority order; and

means for determining the motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for the current block from the candidate set according to the priority order.

12. The apparatus of Claim 1 1 , wherein the priority order is a pre-defined priority order.

13. The apparatus of Claim 12, wherein information associated with the predefined priority order is incorporated in a sequence header, a picture header, or a slice header.

14. The apparatus of Claim 11, wherein the priority order is determined according to an adaptive scheme.

15. The apparatus of Claim 14, wherein the adaptive scheme is based on a criterion selected from a group consisting of:

statistic of reconstructed motion vectors of previous blocks,

partition type of the current block,

correlation of the motion vectors,

directions of motion vectors,

distance of the motion vectors, and

16. The apparatus of Claim 11, wherein the candidate set is selected from said two or more motion vectors.

17. The apparatus of Claim 11, wherein saidmeans for determining the candidate set comprises means for deriving scaled motion vectors corresponding to said two or more motion vectors, wherein the candidate set is selected from the scaled motion vectors or a combination of said two or more motion vectors and the scaled motion vectors.

18. The apparatus of Claim 17, wherein the candidate set is selected from the scaled motion vectors.

19. The apparatus of Claim 17, wherein the candidate set is selected from said two or more motion vectors and the scaled motion vectors.

20. The apparatus of Claim 17,wherein the priority order is according to an adaptive scheme based on whether directions of scaled motion vectorsare interpolation or extrapolation.

21. A method of deriving a motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for a current block in a picture, the method comprising:

receivingtwo or more motion vectors associated with at least one temporal block from a first reference picture list;

arranging the candidate set in a priority order; and

determining the motion vector predictor or motion vector predictor candidateor motion vector or motion vector candidate for the current block from the candidate set according to the priority order;

wherein information associated with the priority order is incorporated in a sequence header, a picture header, or a slice header.

22. The method of Claim 21, wherein the candidate set is selected from said two or more motion vectors.

23. The method of Claim 21, wherein saiddetermining the candidate set comprises deriving scaled motion vectors corresponding to said two or more motion vectors, wherein the candidate set is selected from the scaled motion vectors or a combination of said two or more motion vectors and the scaled motion vectors.

24. The method of Claim 23, wherein the candidate set is selected from the scaled motion vectors.

25. The method of Claim 21, wherein the first reference picture list is a list 0 reference picture or a list 1 reference picture.

26. An apparatus of deriving a motion vector predictor or motion vector predictor candidate ormotion vector or motion vector candidate for a current block in a picture, the apparatus comprising: means for receiving two or more motion vectors associated with at least one temporal block from a first reference pictures list;

means for arranging the candidate set in a priority order; and

means for determining the motion vector predictoror motion vector predictor candidate ormotion vector or motion vector candidate for the current block from the candidate set according to the priority order;

27. The apparatus of Claim 26, wherein the candidate set is selected from said two or more motion vectors.

28. The apparatus of Claim 26, wherein saidmeans for determining the candidate set comprises means for deriving scaled motion vectors corresponding to said two or more motion vectors, wherein the candidate set is selected from the scaled motion vectors or a combination of said two or more motion vectors and the scaled motion vectors.

29. The apparatus of Claim 28, wherein the candidate set is selected from the scaled motion vectors.