Cross component filtering using a temporal source frame
Aspects of the disclosure provide methods and apparatuses for video encoding/decoding. In some examples, an apparatus includes processing circuitry. For example, the processing circuitry obtains reconstructed samples in a temporal source frame that is different from a current frame, and determines filter coefficients of a filter for a current block in the current frame based on at least the reconstructed samples in the temporal source frame. The processing circuitry generates prediction information of the current block based on the determined filter coefficients. The temporal source frame is an inter coded frame that satisfies at least one of a quantization parameter (QP) requirement, a temporal distance requirement, and a reference picture requirement.
1. A method for processing video data, the method comprising:
processing a coded video bitstream that includes the video data according to a format rule, wherein
the coded video bitstream includes prediction information of a current frame and a temporal source frame different from the current frame; and
the format rule specifies that:
a cross component filter is applied to reconstructed samples in the temporal source frame to generate offsets,
the offsets are combined with first reconstructed samples of a current block in the current frame to generate second reconstructed samples of the current block,
the temporal source frame is an inter coded frame; and
at least one of (i) the inter coded frame has a lower quantization parameter (QP) than the current frame, (ii) a temporal distance between the inter coded frame and the current frame is less than a threshold, and (iii) the inter coded frame is a reference picture that is used in motion compensation during a reconstruction of the current frame.
2. The method of claim 1 , wherein the format rule specifies that:
the cross component filter is applied to a luma component of the temporal source frame to generate the offsets for the reconstructed samples of the current block;
the cross component filter is applied to a chroma component of the temporal source frame to generate the offsets for the reconstructed samples of the current block; and
the cross component filter is applied to a weighted sum of the luma component and the chroma component of the temporal source frame to generate the offsets for the reconstructed samples of the current block.
3. The method of claim 1 , wherein the format rule specifies that:
the cross component filter is applied to a first color component of the temporal source frame to generate the offsets for the reconstructed samples of the current block, and
the first color component of the temporal source frame is different from a color component of the reconstructed samples of the current block combined with the offsets.
4. The method of claim 1 , wherein the coded video bitstream includes a syntax element that is signaled in one of a high-level parameter set, a slice header, a tile group header, or a Supplemental Enhancement Information (SEI) message according to the format rule.
5. The method of claim 1 , wherein the format rule specifies that:
motion information of the current block is identified;
a corresponding block in the temporal source frame to the current block is determined based on the identified motion information; and
the cross component filter is applied to at least reconstructed samples of the determined corresponding block to generate the offsets for the reconstructed samples of the current block.
6. The method of claim 5 , wherein the format rule specifies at least one of:
the motion information is derived as a motion vector used for the motion compensation of the current block;
a motion estimation process is performed to identify the motion information; and
the motion estimation process is performed with the motion vector used for the motion compensation of the current block as a starting point to identify the motion information.
7. The method of claim 5 , wherein the format rule specifies at least one of:
a motion estimation process is performed using a luma component of the current block and a luma component of the temporal source frame to identify the motion information;
the motion estimation process is performed using a current color component of the current block and a current color component of the temporal source frame to identify the motion information; and
the motion estimation process is performed using two or more color components of the current block and the two or more color components of the temporal source frame to identify the motion information.
8. The method of claim 5 , further comprising wherein the format rule specifies that:
the cross component filter is applied to a weighed sum of the reconstructed samples of the corresponding block and the reconstructed samples of the current block to generate the offsets.
9. The method of claim 8 , wherein the format rule specifies that:
a second syntax element indicates at least one of a weight applied to the reconstructed samples of the corresponding block or a weight applied to the reconstructed samples of the current block.
10. The method of claim 1 , wherein the inter coded frame has the lower QP than the current frame.
11. The method of claim 1 , wherein the temporal distance between the inter coded frame and the current frame is less than the threshold.
12. The method of claim 1 , wherein the inter coded frame is the reference picture that is used in the motion compensation during the reconstruction of the current frame.
13. A method for video decoding, the method comprising:
generating first reconstructed samples of a current block in a current frame;
obtaining reconstructed samples in a temporal source frame that is different from the current frame;
applying a cross component filter to the reconstructed samples in the temporal source frame to generate offsets; and
combining the first reconstructed samples of the current block with the offsets to generate second reconstructed samples of the current block, wherein
the temporal source frame is an inter coded frame, and
at least one of (i) the inter coded frame has a lower QP value than the current frame; (ii) a temporal distance between the inter coded frame and the current frame is less than a threshold, and (iii) the inter coded frame is a reference picture that is used in motion compensation during a reconstruction of the current frame.
14. The method of claim 13 , wherein the inter coded frame has the lower QP than the current frame.
15. The method of claim 13 , wherein the temporal distance between the inter coded frame and the current frame is less than the threshold.
16. The method of claim 13 , wherein the inter coded frame is the reference picture that is used in the motion compensation during the reconstruction of the current frame.
17. A method for video encoding, the method comprising:
obtaining reconstructed samples in a temporal source frame that is different from a current frame;
applying a cross component filter to the reconstructed samples in the temporal source frame to generate offsets to be applied to first reconstructed samples of a current block in the current frame; and
combining the first reconstructed samples of the current block with the offsets to generate second reconstructed samples of the current block, wherein
the temporal source frame is an inter coded frame, and
at least one of (i) the inter coded frame has a lower QP value than the current frame; (ii) a temporal distance between the inter coded frame and the current frame is less than a threshold, and (iii) the inter coded frame is a reference picture that is used in motion compensation during a reconstruction of the current frame.
18. The method of claim 17 , wherein the inter coded frame has the lower QP than the current frame.
19. The method of claim 17 , wherein the temporal distance between the inter coded frame and the current frame is less than the threshold.
20. The method of claim 17 , wherein the inter coded frame is the reference picture that is used in the motion compensation during the reconstruction of the current frame.