CODING SCHEMES FOR VIRTUAL REALITY (VR) SEQUENCES

Abstract

An improved method for coding video is provided that includes Virtual Reality (VR) sequences that enables more efficient encoding by organizing the VR sequence as a single 2D block structure. In the method, reference picture and subpicture lists are created and extended to account for coding of the VR sequence. To further improve coding efficiency, reference indexing can be provided for the temporal and spatial difference between a current VR picture block and the reference pictures and subpictures for the VR sequence. Further, because the reference subpictures for the VR sequence may not have the proper orientation once the VR sequence subpictures are organized into the VR sequence, reorientation of the reference subpictures is made so that the reference subpicture orientations match the current VR subpicture orientations.

Claims

1. A method of coding a video containing virtual reality (VR) pictures that includes a reference list of past-coded pictures and subpictures, the method comprising: defining a current VR picture in the VR pictures as six subpictures; building at least one reference list for the current VR picture, wherein the at least one reference list holds a reference picture made from a past-coded version of the VR picture as well reference pictures made from past-coded subpictures of the current VR picture; including the past-coded pictures in a first reference list; including the past-coded subpictures in a second reference list; defining motion vector prediction blocks using reference subpictures from the first and second reference list for the current VR picture; and using the motion vector prediction blocks in coding that is sent to the decoder.

2. The method of claim 1, wherein the six subpictures are defined as six tiles within a picture are defined as similar to the concept defined in High Efficiency Video Coding (HEVC).

3. The method of claim 1, further comprising: building a reference index for the current VR picture and subpictures relative to the reference picture and subpictures; indexing the subpictures of the reference picture to the subpictures in the current picture according to temporal and spatial distances to a current block in a current one of the subpictures to a reference block in the reference subpictures; and using the reference list and index in coding that is sent to the decoder.

4. The method of claim 3, wherein for the current block in a current picture, a reference subpicture is assigned a temporal index i and a spatial index j or a combination of temporal and spatial indexes, i+j, wherein the temporal index i can be determined by the temporal distance between the reference picture and the current picture, and wherein the spatial index j can be determined by the spatial distance between the reference subpicture and the current subpicture block.

5. The method of claim 1, further comprising: identifying a current subpicture of the current VR picture; and rotating the subpictures of the reference picture to match the orientation of the subpictures of the current VR picture.

6. A method of coding a video containing virtual reality (VR) pictures that includes indexing of reference subpictures relative to current subpictures, the method comprising: defining a current VR picture in the VR pictures as six subpictures; defining a reference picture and reference subpictures for the current VR picture building a reference list and index for the current VR picture and subpictures relative to the reference picture and subpictures; indexing the subpictures of the reference picture to the subpictures in the current picture according to temporal and spatial distances to a current block in a current one of the subpictures to a reference block in the reference subpictures; and using the reference list and index in coding that is sent to the decoder.

7. The method of claim 6, wherein for the current block in a current picture, a reference subpicture is assigned a temporal index i and a spatial index j or a combination of temporal and spatial indexes, i+j, wherein the temporal index i can be determined by the temporal distance between the reference picture and the current picture, and wherein the spatial index j can be determined by the spatial distance between the reference subpicture and the current subpicture block.

8. The method of claim 6, wherein a closest reference subpicture to the current block in the current subpicture of the current picture temporally and spatially is assigned the index of 0 in the reference picture index, and the second closest reference subpicture is assigned the index of 1 in the reference picture index.

9. The method of claim 6, further comprising: identifying a current subpicture of the current VR picture; and rotating the subpictures of the reference picture to match the orientation of the subpictures of the current VR picture.

10. A method of coding a video containing virtual reality (VR) pictures that includes the ability to change subpicture orientation, the method comprising: defining a current VR picture as six subpictures; identifying subpictures in a reference picture for the current VR picture; identifying a current subpicture of the current VR picture; and rotating the subpictures of the reference picture to match the orientation of the subpictures of the current VR picture.

11. A decoder that provides for temporal motion vector prediction for inter block coding in High Efficiency Video Coding (HEVC) that relies on a block based translational model, the encoder comprising: a processor; and a memory storing code executable by the processor to cause the processor to perform the following steps: identifying a current VR picture in the VR pictures as six subpictures; receiving at least one reference list for the current VR picture, wherein the at least one reference list holds a reference picture made from a past-coded version of the VR picture as well reference pictures made from past-coded subpictures of the current VR picture; wherein the past-coded pictures are included in a first reference list; wherein the past-coded subpictures are included in a second reference list; receiving motion vector prediction blocks using reference subpictures from the first and second reference list for the current VR picture; and using the motion vector prediction blocks in decoding.

12. The decoder of claim 11, wherein the six subpictures are defined as six tiles within a picture are defined as similar to the concept defined in High Efficiency Video Coding (HEVC).

13. The decoder of claim 11, wherein the memory further stores code to cause the processor to perform the following additional steps: receiving a reference index for the current VR picture and subpictures relative to the reference picture and subpictures; receiving an index of the subpictures of the reference picture to the subpictures in the current picture according to temporal and spatial distances to a current block in a current one of the subpictures to a reference block in the reference subpictures; and using the reference list and index in decoding.

14. The method of claim 13, wherein for the current block in a current picture, a reference subpicture is assigned a temporal index i and a spatial index j or a combination of temporal and spatial indexes, i+j, wherein the temporal index i can be determined by the temporal distance between the reference picture and the current picture, and wherein the spatial index j can be determined by the spatial distance between the reference subpicture and the current subpicture block.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Further details of the present invention are explained with the help of the attached drawings in which:

[0017] FIG. 1 illustrates how for a VR sequence a 360 degree spherical object can be mapped onto surfaces of a cube;

[0018] FIG. 2 illustrates the numbered surfaces of the cube that has been mapped with a VR sequence from an internal spherical structure;

[0019] FIG. 3 shows organization of the cube surfaces of FIG. 2 into a 4×3 two dimensional structure for coding of the VR sequence;

[0020] FIG. 4 shows organization of the cube surfaces of FIG. 2 into a 3×2 two dimensional structure for coding of the VR sequence;

[0021] FIG. 5 provides a flowchart with steps according to embodiments of the present invention for coding video using reference picture and subpicture lists to account for a VR sequence;

[0022] FIG. 6 provides a flowchart with steps indicating how reference subpicture indexing is provided according to embodiments of the present invention;

[0023] FIG. 7 illustrates pictures used to create a reference list index with a reference subpicture assigned a temporal index, i, and a spatial index, j;

[0024] FIG. 8 shows how the six subpictures in a reference picture are rotated for a current subpicture 2;

[0025] FIG. 9 shows how subpictures of a reference picture are rotated to have the same orientation with the current picture ranging from subpicture 0 through 5;

[0026] FIG. 10 provides a flowchart with steps showing how VR reference subpicture orientation is changed so that the orientation matches the current subpicture; and

[0027] FIG. 11 shows an encoder and decoder that can be configured to perform encoding and decoding with VR pictures according to embodiments of the present invention.

DETAILED DESCRIPTION

[0028] A VR sequence in a video can be treated as a regular 2D sequence with six subpictures for the embodiments of the invention described herein. That is, each VR picture is treated as a single 2D picture and coding standards such as AVC and HEVC can be applied to the single VR sequence directly. The VR picture can be a 4×3 or 3×2 breakdown of a cube into six subpictures at each time instance, as illustrated in FIGS. 3 and 4. The six VR picture subpictures can be treated as six tiles within a picture, similar to the concept defined in HEVC.

[0029] To accomplish motion estimation and compensation (ME/MC) for embodiments of the present invention, the concept of reference pictures lists, reference indexing and an orientation of references relative to a current picture can be provided for a VR sequence for embodiments of the present invention. A description of each of these concepts is provided to follow.

A. Reference Lists

[0030] The concept of reference pictures and lists can be extended for a VR sequence. Similar to AVC and HEVC, for a block in a current subpicture within a current picture, reference pictures can be provided and reference lists built to enable ME/MC. Reference pictures can be built from the past-coded pictures of subpictures as well as the past-coded subpictures of the current picture. A listing of these reference pictures can further be created.

[0031] The past-coded pictures can be included in at least one reference list, similar to AVC and HEVC. The past-coded subpictures for the current picture may be included in a second reference list.

[0032] Now for blocks, consider a current block in a current subpicture within a current picture. For the current block the reference prediction block can be found in one of the reference subpictures per reference list. One of reference subpictures in which the reference prediction block is found can be in one of the past-coded pictures in a different picture time instance than the current time instance forming the reference.

[0033] FIG. 5 provides a flowchart with steps according to embodiments of the present invention for coding video using reference picture and subpicture lists to account for a VR sequence. In a first step 500, the method defines a current VR picture provided in a video to have six subpictures. Next in a step 502, at least one reference list for the current VR picture is built, wherein the at least one reference list holds a past-coded version of the VR picture as well as the past-coded subpictures of the current VR picture. In step 504, the past-coded pictures are separated out into a first reference list. In step 506, the past-coded subpictures are included in a second reference list. In step 508, motion vector prediction blocks are defined using the reference subpictures from the first and second reference list for the current VR picture. Finally, in step 510 the motion vector prediction blocks are used in coding that is sent to the decoder.

B. Reference Indexing

[0034] Due to the fact that the closer the reference picture and subpictures are to the current subpicture temporally and spatially, the higher the correlation between the reference picture and subpictures and the current picture, the reference pictures and subpictures for embodiments of the present invention may be indexed according to their temporal and spatial distance to the current subpicture.

[0035] Embodiments of the present invention provide for a default reference picture/subpicture index order. In particular, for a current block in a current subpicture for a current picture, a reference picture and subpictures in a reference picture list are indexed according to its temporal and spatial distances to the current block in the current subpicture of the current picture. In other words, the closest reference picture/subpicture to the current block in the current subpicture of the current picture temporally and spatially is assigned the index of 0, the second closest reference picture/subpicture is assigned the index of 1, and so on.

[0036] FIG. 6 provides a flowchart with steps indicating how reference subpicture indexing is provided according to embodiments of the present invention. In particular, the method illustrated by the flowchart of FIG. 6 provides for coding a video containing virtual reality (VR) pictures that includes indexing of reference subpictures relative to current subpictures. In a first step 600, a current VR picture in the VR pictures is defined to include six subpictures. Next in step 602, a reference picture and reference subpictures for the current VR picture is defined. In step 604 a reference list and index is built for the current VR picture and subpictures relative to the reference picture and subpictures. In step 606, indexing of the subpictures of the reference picture to the subpictures in the current picture is provided according to temporal and spatial distances to a current block in a current one of the subpictures relative to a reference block in the reference subpictures. Finally, in step 608, the reference list and index are used in coding that is sent to the decoder.

[0037] In embodiment for providing a reference list index, a reference subpicture is assigned a temporal index, i, and a spatial index, j, or a combination of temporal and spatial indexes, i+j. The temporal index, i, can be determined by the temporal distance between the reference picture and the current picture, i.e., the closer, the smaller the index. The spatial index, j, can be determined by the spatial distance between the reference subpicture in the reference picture and the current block collocated in the reference picture.

[0038] FIG. 7 illustrates pictures used to create a reference list index with a reference subpicture assigned a temporal index, i, and a spatial index, j. In FIG. 7, a current block 702 in gray color is shown in subpicture 0 of a current picture 700. As seen, in the middle of FIG. 7, the closest subpicture to the collocated block 712 of the current block in a reference picture 710 is subpicture 2. Hence, for the current block, subpicture 2 in any reference picture of any reference list will be assigned a spatial reference index of j=0. Subpicture 1 is the second closet subpicture, and so, it will be assigned the spatial reference index of j=1. For this example, for the current block in subpicture 0 of the current picture, the spatial reference indexes of j=0, 1, 2, 3, 4, and 5 will respectively be assigned to subpictures 2, 1, 4, 3, and 5 of any reference picture of any reference list.

C. Subpicture Rotation

[0039] Not all the subpictures in a reference picture have the same orientation as the current subpicture of a current VR picture. To enable coding of the VR picture efficiently, the orientation of the six subpictures making up the VR picture that is made up of arranged faces of a cube should be organized to have the same orientation irrespective of arrangement of the cube faces. FIG. 8 shows how the six subpictures in a reference picture are rotated for a current subpicture 2. A seen, in this example, subpicture 1 needs to be rotated by 90 degree counterclockwise, subpicture 4 to be rotated 90 degree counterclockwise and subpicture 5 needs to be rotated by 180 degree clockwise. FIG. 9 shows how subpictures of a reference picture are rotated to have the same orientation with the current subpicture ranging from picture 0 through 5.

[0040] Accordingly, embodiments of the present invention provide for the subpictures of a reference picture to be rotated as shown in FIG. 9 accordingly so that they can have the same orientation as the current subpicture, before any prediction is performed. FIG. 10 provides a flowchart with steps showing how VR reference subpicture orientation is changed so that the orientation matches the current subpicture. In a first step, 1000, a current VR picture in the VR pictures is defined to include six subpictures. Next in step 1002, subpictures for a reference picture for the current VR picture are identified. In step 1004 a current subpicture of the current VR picture is identified. Finally, in step 1006 subpictures of the reference picture are oriented to match the orientation of the current subpicture of the current VR picture.

[0041] For better temporal and spatial prediction, the subpictures in a reference picture are rotated and rearranged accordingly so that the spatial content transition from a subpicture to its neighbor subpictures within the reference picture can be continuous and smooth. It is noted that in addition with rotation so that arrangement of subpictures of the current and reference pictures are the same, the spatial reference index, j, may not be necessary as the reference picture of six subpictures can be treated as one single picture in the reference list.

[0042] FIG. 11 shows an encoder 102 and decoder 104 that can be configured to perform encoding and decoding with VR pictures according to embodiments of the present invention. Motion estimation and motion compensation is performed using information from embodiments of the present invention with encoder 102 and decoder 1104 using a process of determining a motion vector (MV) for a current unit of video. For example, the motion estimation process searches for a best match prediction for a current unit block of video (e.g., a prediction block) over reference pictures. Motion compensation is then performed by subtracting a reference unit pointed to by the motion vector from the current unit of video.

[0043] To perform motion estimation and compensation, encoder 1102 and decoder 1104 include motion estimation and compensation blocks 1104-1 and 1104-2, respectively. For bi-directional prediction, the motion estimation and compensation blocks 1104-1 and 1104-2 can use a combined bi-directional reference unit in the motion compensation process for the current unit.

[0044] For the encoder 1102 and decoder 1104 of FIG. 11, embodiments of the present invention contemplate that software to enable them to perform functions described to follow for the present invention is provided in a memory. The encoder 1102 and decoder 1104 are further contemplated to include one or more processors that function in response to executable code stored in the memory to cause the processor to perform the functions described.

[0045] Although the present invention has been described above with particularity, this was merely to teach one of ordinary skill in the art how to make and use the invention. Many additional modifications will fall within the scope of the invention as that scope is defined by the following claims.