METHOD FOR DERIVING A MOTION VECTOR

Abstract

A method for decoding video includes creating a first list of motion vectors from at least one neighboring block in a current frame of the video and creating a second list of motion vectors from at least one previous block in a temporally earlier frame of the video. A third list of motion vectors is created based upon the first list and the second list. Based upon receiving a motion vector competition control parameter selecting one of the motion vectors from the third list, wherein the second list of motion vectors is further based upon a flooring function.

Description

BACKGROUND OF THE INVENTION

The present invention relates to image decoding using buffer compression for motion vector competition.

Existing video coding standards, such as H.264/AVC, generally provide relatively high coding efficiency at the expense of increased computational complexity. As the computational complexity increases, the encoding and/or decoding speeds tend to decrease. Also, the desire for increasingly accurate motion estimation tends to require increasingly larger memory requirements. The increasing memory requirements tend to result in increasingly more expensive and computationally complex circuitry, especially in the case of embedded systems.

The foregoing and other objectives, features, and advantages of the invention will be more readily understood upon consideration of the following detailed description of the invention, taken in conjunction with the accompanying drawings.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

FIG. 1 illustrates an encoder and decoder.

FIG. 2 illustrates a vector decoding process.

FIG. 3 illustrates a partial pair of two frames.

FIG. 4 illustrates a motion vector competition process.

FIG. 5 illustrates a particular motion vector competition process.

FIG. 6 illustrates a decoder with motion vector competition.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENT

Referring to FIG. 1, a video system typically includes a video source such as an encoder or storage for encoded video, and a decoder. In the video decoder, one of the aspects that facilitates higher compression performance for high resolution video is larger block structures with flexible mechanisms of sub-partitioning. For example, the TMuC (JCT-VC, “Test Model under Consideration”, document JCTVC-A205 of JCT-VC, (2010), incorporated by reference herein) defines coding units (CUs) which define sub-partitioning of an image into rectangular regions. The coding unit contains one or more prediction units (PUs) and transform units (TUs). The partition geometry of these elements may be encoded using a tree segmentation structure. At the level of the PU, either intra-picture or inter-picture prediction may be selected. Intra-picture prediction may be performed from samples of already decoded adjacent PUs, where the different modes may include (1) DC (flat average), (2) horizontal, (3) vertical, (4) one of up to 28 angular directions (number depending on block size), (5) plane (amplitude surface) prediction, and (6) bilinear prediction. The signaling of the mode is derived from the modes of adjacent PUs. Inter-picture prediction is performed from regions(s) of already decoded pictures stored in the reference picture buffer. This permits selection among multiple reference pictures, as well as bi-prediction (including weighted averaging) from two reference pictures or two positions in the same reference picture. The reference area is selected by specifying a motion vector displacement and a reference picture index, which may generally be referred to herein as a motion vector. For efficient encoding, derivation of motion vectors from those of adjacent PUs may be made by a median computation or as motion vector competition. The motion vector difference between the predicted motion vectors and the actual motion vectors may also be transmitted. Motion compensation may be performed with a motion vector precision up to quarter-sample or one-eights sample precision.

Referring to FIG. 2, motion estimation for the current image uses a motion vector coding process that generally includes receiving a motion vector prediction control parameter in the bit stream 100. The motion vector prediction control parameter may be an index value used to select from among a group of available motion vectors. A motion vector prediction 110 may be generated based upon the motion vector prediction control parameter 100. The decoder may also receive a motion vector difference 120 in the bit stream. The motion vector difference 120 and the predicted motion vector 110 are used to generate a motion vector for the current image 130. In general, the bit stream provided to the decoder includes, among other data, a control parameter for the motion vector, and the difference between the prediction generated by the motion vector competition tool and the desired motion vector.

Referring to FIG. 3, for the motion vector competition process a list of candidate motion vectors may be created. In general, the potential motion vectors, may include for example, a motion vector MVB from above the current block A, a motion vector MVC to the left of the current block A, and a motion vector MVD to the above left of the current block A. The potential motion vectors may, for example, include one or more of all of the adjacent blocks. The list of candidate motion vectors may also include one or more motion vectors from one or more previously encoded frames. For example, one of the previous motion vectors may be the co-located block A′ of the previous frame with the current block A. It is noted that often the video coding and decoding is not performed in the same temporal order in which it is displayed.

Referring to FIG. 4, a particular implementation of a motion vector competition process utilizes a list of motion vector(s) of neighboring block(s) in the current frame 400 and a list of motion vector(s) from previous block(s) in a previously transmitted frame 410. The two lists of motion vectors 400, 410 are merged 420 to create a single list. Duplicate values in the list are removed 430. The motion vector competition control parameter 440 is received, which is typically an index. The motion vector competition control parameter, or equivalently the motion vector control parameter, 440 is used to select 450 an appropriate motion vector in the list 430 at the location indicated. The selected motion vector 450 is provided as the output selected motion vector 460. The motion vector difference 470 between the predicted motion vector and the actual motion vector may also be provided. Thus, in a block by block manner, the motion vector competition tool predicts the motion vector for the current block. In general, the motion vector or any data which may be used to determine the motion vector may be transmitted.

As an illustrative example, referring to FIG. 5, a motion vector competition control parameter (MVCCP) is received in the bit stream, such as for example 3. The motion vector competition control parameter indicates where in the list of candidate motion vectors the desired motion vector is. If a set of motion vectors MVB, MVC, MVD, MVA′ is created then the motion vector control parameter would indicates that the third motion vector MVD should be selected for the current block A as the motion vector prediction 460. For example, MVCCP 1 may be MVB, MVCCP 2 may be MVC, MVCCP 3 may be MVD, and MVCCP 4 may be MVA′.

In particular to creating the list of motion vectors from blocks in previously transmitted frames 410, several techniques have been previously used. For example, one such technique includes for a block in the current frame that spans the pixel locations (x,y) to (x+N, y+M), the motion vector from the previous frame used for the pixel prediction located at (x,y) is selected to be included in the list. For example, x,y may indicate the top-left position of the block, N may indicate the horizontal size of the block, and M may indicate the vertical size of the block. For example, another such technique includes for a block in the current frame that spans the pixel locations (x,y) to (x+N, y+M), the motion vector from the previous frame used for the pixel prediction located at (x+N/2, y+M/2) may be selected to be included in the list. While such techniques may select desirable motion vectors, the resulting list of motion vectors may tend to be large, resulting in significant memory requirements.

A modified technique for the current block in the current frame that spans the pixel locations (x,y) to (x+N, y+M) is to select and to store in the memory buffer the motion vector from the previous frame used for the pixel prediction located at ((x>>Z)<<Z, (y>>Z)<<Z) where Z is an integer. Preferably Z is 4. It is noted that (x>>Z)<<Z is generally equivalent to “the floor of (x/(2{circumflex over ( )}Z))” *(2{circumflex over ( )}Z). A floor function maps a real number to the largest previous integer value. For example, the floor (x)=[X] is the largest integer not greater than x. This technique is effective for a block size of 4×4 or, alternatively, 16×16.

Another modified technique for the current block in the current frame that spans pixels locations (x,y) to (x+N, y+M) is to select and to store in the memory buffer the motion vector from the previous frame used for the pixel prediction located at (((x>>Z)<<Z+Y, ((y>>Z)<<Z)+Y), where Y and Z are integers. Preferably Z is 4 and Y is 8. Also, Y may be 1<< (Z−1), which is generally equivalent to (2{circumflex over ( )}Z)/2. It is noted that ((x>>Z)<<Z)+Y is generally equivalent to “the floor of (x/(2{circumflex over ( )}Z))” *(2{circumflex over ( )}Z)+Y. This technique is effective for a block size of 4×4 or, alternatively, 16×16.

Yet another modified technique for the current block in the current frame that spans pixel location (x,y) to (x+N, y+M) is to select and to store in the memory buffer the motion vector from the previous frame used for the pixel prediction located at ((((x+N/2)>>Z)<<Z)+Y,(((y+M/2)>>Z)<<Z)+Y), where Y and Z are integers. Preferably Z is 4 and Y is 8. Also, Y may be 1<< (Z-1) which is generally equivalent to (2{circumflex over ( )}Z)/2. It is noted that (((x>>Z)<<Z)+Y is generally equivalent to “the floor (x/(2{circumflex over ( )}Z))” *(2{circumflex over ( )}Z)+Y. This technique is effective for a block size of 4×4 or, alternatively, 16×16.

A further modified technique for the current block in the current frame that spans pixel location (x,y) to (x+N, y+M) is to select and to store in the memory buffer the motion vector from the previous frame used for the pixel prediction located at ((((x+N/2)>>Z)<<Z)+Y, (((y+M/2)>>Z)<<Z)+Y), where Y and Z are integers. Preferably Z is 4 and Y is 4. Also, Y may be 1<< (Z-1) which is generally equivalent to (2{circumflex over ( )}Z)/2. It is noted that (((x>>Z)<<Z)+Y is generally equivalent to “the floor (x/(2{circumflex over ( )}Z))” *(2{circumflex over ( )}Z)+Y. This technique is effective for a block size of 4×4 or, alternatively, 16×16.

In yet another modified technique for the current block in the current frame that spans pixel locations (x,y) to (x+N,y+M) is where the motion vector from the previous frame for the co-located block is coded by interpolation prediction, and the system may select among multiple motion vectors belonging to the co-located block. In this embodiment, a first motion vector for the current block is determined using a first motion vector in the previous frame used for the pixel prediction located at ((((x+N/2)>>Z)<<Z)+Y, (((y+M/2)>>Z)<<Z)+Y), where Y and Z are integers. Additionally, a second motion vector for the current block is determined using a second motion vector in the previous frame used for the pixel prediction located at ((((x+N/2)>>Z)<<Z)+Y, (((y+M/2)>>Z)<<Z)+Y), where Y and Z are integers. Preferably Z is 4 and Y is 4. Also, Y may be 1<< (Z−1) which is generally equivalent to (2{circumflex over ( )}Z)/2. It is noted that (((x>>Z)<<Z)+Y is generally equivalent to “the floor (x/(2{circumflex over ( )}Z))” *(2{circumflex over ( )}Z)+Y. This technique is effective for a block size of 4×4 or, alternatively, 16×16.

Modified techniques may be used to select and to store in the memory buffer the motion vector from a previous frame which no longer requires the storage of every motion vector used in the prediction of previous frames. In a preferred embodiment, the motion vectors that are not used by the modified techniques are removed from memory after all steps required to decode the current frame are completed. In an alternative embodiment, the motion vectors that are not used by the modified techniques are removed from memory after being used for prediction but before all steps required to the decode the current frame are completed. In a specific example, the motion vectors that are not used by the modified techniques are removed from memory before a de-blocking operation is performed, and the de-blocking operation uses only the motion vectors that are not used by the modified techniques for processing.

Referring to FIG. 6, a motion vector competition technique may likewise include a decoder that receives a bit stream from which a motion vector competition is applied. This data may include any suitable data previously described. In some embodiments, the motion vector competition technique that includes a buffer compression may be used, without necessarily including the additional features described in relation to FIG. 4 and FIG. 5.

By using a flooring function for the selection of candidate vectors from the previous frame the number of candidate vectors from the previous frame is reduced. In this manner, the system may effectively reduce the number of vectors without likely eliminating the most appropriate motion vector from the available selection. Moreover, by reducing the list of candidate vectors in an effective manner, the memory buffer may be reduced, which reduces the complexity of the system, especially suitable for embedded systems.

In particular to creating the list of motion vectors from blocks in previously transmitted frames 410, typically the motion vector is only added to the list if it is available. Thus, for a current block, one the system only uses the motion vector from a previous frame if the co-located block uses a motion vector. While functional, such an approach does not take into account the full nature of the video stream of encoded frames.

A modified technique for the current block in the current frame that spans the pixel locations (x,y) to (x+N, y+M) is to only to select and to store in the memory buffer a motion vector if the pixel at location ((x>>Z)<<Z, (y>>Z)<<Z) in the previous frame is coded using an inter-coding technique. Z is preferably an integer. In general, this means that the location in a previous frame can not be coded with an intra-prediction technique.

Another modified technique for the current block in the current frame that spans pixels locations (x,y) to (x+N, y+M) is to only to select and to store in the memory buffer a motion vector if the pixel at location ((x>>Z)<<Z+Y, (y>>Z)<<Z)+Y) in the previous frame is coded using an inter-coding technique. Preferably Y and Z are integers. In general, this means that the location in a previous frame can not be coded with an intra-prediction technique.

Yet another modified technique for the current block in the current frame that spans pixel location (x,y) to (x+N, y+M) is to only to select and to store in the memory buffer a motion vector if the pixel at location ((x+N/2>>Z)<<Z)+Y, (y+M/2>>Z)<<Z)+Y) in the previous frame is coded using an inter-coding technique. Preferably Y and Z are integers. In general, this means that the location in a previous frame can not be coded with an intra-prediction technique.

In a preferred implementation a combination of the flooring function is combined with the limitation of an inter-coding technique. For example, the system may only select a motion vector from a previous frame if the co-located block is coded using an inter-coding technique. Furthermore, preferably if the system points to an intra-coded block in the previous frame, then the motion vector may be assumed to be 0 at that location.

The terms and expressions which have been employed in the foregoing specification are used therein as terms of description and not of limitation, and there is no intention, in the use of such terms and expressions, of excluding equivalents of the features shown and described or portions thereof, it being recognized that the scope of the invention is defined and limited only by the claims which follow.

Claims

1. A method of motion vector prediction corresponding to a video bitstream, the method comprising: determining, based at least in part on coordinates of a current block, a plurality of locations, each corresponding to a block in a previous frame associated with the current block, the determining including performing an arithmetic right shift operation;obtaining, based at least in part on the plurality of locations, a plurality of motion vector candidates corresponding to the blocks in the previous frame;selecting a motion vector from the plurality of motion vector candidates; andpredicting the current block based at least in part on the selected motion vector.

2. The method of claim 1, wherein the coordinates of the current block comprise top-left coordinates of the current block, and wherein the determining of the plurality of locations further comprises: dividing a horizontal size of the current block by two and adding the divided horizontal size to an x-coordinate of the top-left coordinates; anddividing a vertical size of the current block by two and adding the divided vertical size to a y-coordinate of the top-left coordinates.

3. The method of claim 1, further comprising: deriving at least one neighboring block motion vector candidate corresponding to at least one neighboring block of the current block, wherein the plurality of motion vector candidates included for the selecting of the motion vector further comprises the at least one neighboring block motion vector candidate.

4. The method of claim 3, wherein the at least one neighboring block is adjacently located above the current block, above left of the current block, or left of the current block.

5. The method of claim 1, wherein the determining of the plurality of locations further includes performing an arithmetic left shift operation on a result obtained from the arithmetic right shift operation, wherein the arithmetic right shift operation and the arithmetic left shift operation have a same magnitude.

6. The method of claim 5, wherein the same magnitude is 4.

7. A non-transitory computer-readable storage medium having stored thereon instructions executable by processing circuitry to perform operations for motion vector prediction corresponding to a video bitstream, the operations comprising: determining, based at least in part on coordinates of a current block, a plurality of locations, each corresponding to a block in a previous frame associated with the current block, the determining including performing an arithmetic right shift operation;obtaining, based at least in part on the plurality of locations, a plurality of motion vector candidates corresponding to the blocks in the previous frame;selecting a motion vector from the plurality of motion vector candidates; andpredicting the current block based at least in part on the selected motion vector.

8. The non-transitory computer-readable storage medium of claim 7, wherein the coordinates of the current block comprise top-left coordinates of the current block, and wherein the determining of the plurality of locations further comprises: dividing a horizontal size of the current block by two and adding the divided horizontal size to an x-coordinate of the top-left coordinates; anddividing a vertical size of the current block by two and adding the divided vertical size to a y-coordinate of the top-left coordinates.

9. The non-transitory computer-readable storage medium of claim 7, the operations further comprising: deriving at least one neighboring block motion vector candidate corresponding to at least one neighboring block of the current block, wherein the plurality of motion vector candidates included for the selecting of the motion vector further comprises the at least one neighboring block motion vector candidate.

10. The non-transitory computer-readable storage medium of claim 9, wherein the at least one neighboring block is adjacently located above the current block, above left of the current block, or left of the current block.

11. The non-transitory computer-readable storage medium of claim 7, wherein the determining of the plurality of locations further includes performing an arithmetic left shift operation on a result obtained from the arithmetic right shift operation, wherein the arithmetic right shift operation and the arithmetic left shift operation have a same magnitude.

12. The non-transitory computer-readable storage medium of claim 11, wherein the same magnitude is 4.

13. A device for performing motion vector prediction corresponding to a video bitstream comprising: at least one processor; andmemory having instructions which, when executed by the at least one processor, cause the device to perform operations comprising: determining, based at least in part on coordinates of a current block, a plurality of locations, each corresponding to a block in a previous frame associated with the current block, the determining including performing an arithmetic right shift operation;obtaining, based at least in part on the plurality of locations, a plurality of motion vector candidates corresponding to the blocks in the previous frame;selecting a motion vector from the plurality of motion vector candidates; andpredicting the current block based at least in part on the selected motion vector.

14. The device of claim 13, wherein the coordinates of the current block comprise top-left coordinates of the current block, and wherein the determining of the plurality of locations further comprises: dividing a horizontal size of the current block by two and adding the divided horizontal size to an x-coordinate of the top-left coordinates; anddividing a vertical size of the current block by two and adding the divided vertical size to a y-coordinate of the top-left coordinates.

15. The device of claim 13, the operations further comprising: deriving at least one neighboring block motion vector candidate corresponding to at least one neighboring block of the current block, wherein the plurality of motion vector candidates included for the selecting of the motion vector further comprises the at least one neighboring block motion vector candidate.

16. The device of claim 15, wherein the at least one neighboring block is adjacently located above the current block, above left of the current block, or left of the current block.

17. The device of claim 13, wherein the determining of the plurality of locations further includes performing an arithmetic left shift operation on a result obtained from the arithmetic right shift operation, wherein the arithmetic right shift operation and the arithmetic left shift operation have a same magnitude.

18. The device of claim 17, wherein the same magnitude is 4.