Apparatus and method of constrained partition size for high efficiency video coding
An apparatus and method for video coding and decoding with constrained PU partition are disclosed. In the High Efficient Video Coding (HEVC) system, rate-distortion function or other performance criterion usually is evaluated for various CU partition and PU partition during the encoding process in order to select a configuration with best possible performance. The PU design in the current HEVC development results in some redundancy that causes rate-distortion function or other performance criterion repeatedly evaluated for same PU configuration. Accordingly, constrained PU partition is developed to eliminate or reduce the redundancy in processing. Furthermore, necessary syntax to convey the information related to constrained PU partition between an encoder and a decoder is developed. Systems embodying the present invention has been shown to result in sizeable reduction in encoding and decoding time while the performance in terms of RD-rate remains approximately the same or slightly higher than a conventional HEVC system.
1. A method of video processing for a video encoder, the method comprising:
receiving video data corresponding to a 2N×2N leaf CU (coding unit) at depth k, wherein N is a first integer greater than 0 and k is a second integer greater than or equal to 0;
partitioning the video data corresponding to the 2N×2N leaf CU into a set of PUs (prediction units) according to a prediction mode of the 2N×2N leaf CU, wherein the set includes N×N partition size and the prediction mode includes an INTER mode and an INTRA mode;
removing the N×N partition size for the prediction mode if the 2N×2N leaf CU is larger than SCU (smallest CU), wherein a constrained set of PUs is formed when the partition size is removed; and
applying prediction on the video data according to the prediction mode to generate residue data, wherein said applying prediction on the video data is based on the set of PUs if the 2N×2N leaf CU corresponds to the SCU and said applying prediction on the video data is based on the set of constrained PUs if the 2N×2N leaf CU is larger than the SCU.
2. The method of claim 1 , wherein the set of PUs further comprises 2N×2N partition size, 2N×N partition size and N×2N partition size for the INTER mode.
3. The method of claim 1 , wherein the set of PUs further comprises 2N×2N partition size for the INTRA mode.
4. The method of claim 1 , further comprising a step of incorporating a flag to indicate whether any of next coding units at a next depth is further split.
5. The method of claim 4 , wherein if none of the next coding units at the next depth is further split as indicated by the flag, no further information is incorporated to indicate split decision for each of the next coding units at the next depth.
6. A method of video processing for a video encoder, the method comprising:
receiving video data corresponding to a 2N×2N leaf CU (coding unit) at depth k, wherein N is an integer greater than 0;
for each N×N leaf CU of the 2N×2N leaf CU:
evaluating condition 1 to determine if k>0;
evaluating condition 2 to determine if said each N×N leaf CU is a fourth leaf CU of the 2N×2N leaf CU;
evaluating condition 3 to determine if three prior leaf CUs have the depth k; and
evaluating condition 4 to determine if said three prior leaf CUs have a mode of same type and have N×N partition size;
removing 2N×2N partition size to form a constrained partition set if the condition 1 through the condition 4 are all asserted;
partitioning said each N×N leaf CU according to the constrained partition set if the condition 1 through the condition 4 are all asserted; and
applying prediction on the video data to generate residue data based on the constrained partition set.
7. The method of claim 6 , wherein the mode type is INTER.
8. The method of claim 6 , wherein the mode type is INTRA.
9. A method for decoding of a video bitstream associated with video data in a video decoder, wherein the video data comprises a 2N×2N leaf CU (coding unit) at depth k partitioned into a set of PUs (prediction units) according to a prediction mode of the 2N×2N leaf CU, wherein the set includes N×N partition size and the prediction mode includes an INTER mode and an INTRA mode, the method comprising:
performing a test to determine if the 2N×2N leaf CU is larger than SCU (smallest CU);
selecting codewords for the set of PUs according to the test;
decoding the video bitstream to recover PU partition according to the codewords selected;
decoding the video bitstream to recover residue data associated with the 2N×2N leaf CU; and
reconstructing video data associated with the 2N×2N leaf CU based on the set of PUs using the residue data associated with the 2N×2N leaf CU.
10. The method of claim 9 , further comprising:
extracting a flag from the video bitstream, wherein the flag indicates whether any of next coding units at a next depth is further split; and
recovering structure of the next coding units at the next depth according to the flag.
11. The method of claim 10 , wherein if none of the next coding units at the next depth is further split as indicated by the flag, no further information is incorporated in the bitstream for recovering structure of the next coding units at the next depth.
12. An apparatus for video coding, the apparatus comprising:
means for receiving video data corresponding to a 2N×2N leaf CU (coding unit) at depth k, wherein N is a first integer greater than 0 and k is a second integer greater than or equal to 0;
means for partitioning the video data corresponding to the 2N×2N leaf CU into a set of PUs (prediction units) according to a prediction mode of the 2N×2N leaf CU, wherein the set includes N×N partition size and the prediction mode includes an INTER mode and an INTRA mode;
means for removing the N×N partition size for the prediction mode if the 2N×2N leaf CU is larger than SCU (smallest CU), wherein a constrained set of PUs is formed when the partition size is removed; and
means for applying prediction on the video data according to the prediction mode to generate residue data, wherein said applying prediction on the video data is based on the set of PUs if the 2N×2N leaf CU corresponds to the SCU and said applying prediction on the video data is based on the set of constrained PUs if the 2N×2N leaf CU is larger than the SCU.
13. The apparatus of claim 12 , wherein the set of PUs further comprises 2N×2N partition size, 2N×N partition size and N×2N partition size for the INTER mode.
14. The apparatus of claim 12 , wherein the set of PUs further comprises 2N×2N partition size for the INTRA mode.
15. The apparatus of claim 12 , further comprising means for incorporating a flag when the set of PUs contains more than one PU to indicate whether any of said more than one PU is further split.
16. An apparatus for video processing, the apparatus comprising:
means for receiving video data corresponding to a 2N×2N leaf CU (coding unit) at depth k, wherein N is an integer greater than 0;
for each N×N leaf CU of the 2N×2N leaf CU:
means for evaluating condition 1 to determine if k> 0 ;
means for evaluating condition 2 to determine if said each N×N leaf CU is a fourth leaf CU of the 2N×2N leaf CU;
means for evaluating condition 3 to determine if three prior leaf CUs have the depth k; and
means for evaluating condition 4 to determine if said three prior leaf CUs have a mode of same type and have N×N partition size;
means for removing 2N×2N partition size if the condition 1 through the condition 4 are all asserted;
means for partitioning said each N×N leaf CU according to a constrained partition set if the condition 1 through the condition 4 are all asserted; and
means for applying prediction on the video data to generate residue data based on the constrained partition set.
17. The apparatus of claim 16 , wherein the mode type is INTER.
18. The apparatus of claim 16 , wherein the mode type is INTRA.
19. A apparatus for decoding of a video bitstream associated with video data, wherein the video data comprises a 2N×2N leaf CU (coding unit) at depth k partitioned into a set of PUs (prediction units) according to a prediction mode of the 2N×2N leaf CU, wherein the set includes N×N partition size and the prediction mode includes an INTER mode and an INTRA mode, the apparatus comprising:
means for performing a test to determine if the 2N×2N leaf CU is larger than SCU (smallest CU);
means for selecting codewords for the set of PUs according to the test;
means for decoding the video bitstream to recover PU partition according to the codewords selected;
means for decoding the video bitstream to recover residue data associated with the 2N×2N leaf CU; and
means for reconstructing video data associated with the 2N×2N leaf CU based on the set of PUs using the residue data associated with the 2N×2N leaf CU.
20. The apparatus of claim 19 , further comprising:
means for extracting a flag from the video bitstream , wherein the flag indicates whether any of next coding units at a next depth is further split; and
means for recovering structure of the next coding units at the next depth according to the flag.