Video data prediction method and device

Abstract

A method and apparatus for video encoding and decoding using a sub-block-based affine mode are disclosed. According to the method, for a block, a control point motion vector (MV) associated with the affine mode is determined. The sub-block MV of a target sub-block of the block is derived from the control point MV of the block. Based on prediction refinement with optical flow (PROF), a prediction offset of a target pixel of the target sub-block is determined using information including a pixel MV offset of the sub-block MV relative to the target pixel. The target pixel of the target sub-block is encoded or decoded using a modified predictor. A modified predictor is generated by clipping the prediction offset to a target range and combining the clipped prediction offset with the original predictor.

Claims

1. A method for predicting video data, performed by a video encoder or a video decoder using a set of codec tools including an affine mode, the method comprising: receiving input data related to a current block at a video encoder side, or receiving a video bitstream corresponding to compressed data including the current block at a video decoder side, wherein the affine mode is allowed for the current block; determining two or more control point motion vectors associated with the affine pattern for the current block; deriving a sub-block motion vector of a target sub-block of the current block from the two or more control point motion vectors of the current block; deriving a prediction offset of the target pixel of the target sub-block from the sub-block motion vector of the target pixel using first information including a pixel motion vector offset according to prediction refinement with optical flow; as well as encoding or decoding the target pixel of the target subblock using a modified predictor, wherein the modified predictor is generated by clipping the prediction offset to a first target range and combining the clipped prediction offset with an original predictor, or the modified predictor is generated by clipping a refined predictor to a second target range, and wherein the refined predictor is generated by combining the original predictor and the prediction offset, where the first target range corresponds to [–dILimit, dILimit–1], where dILimit is set to 1<<max(13, bit depth + 1), and the bit depth corresponds to the sample bit depth of the current block; the second target range corresponds to [–2 (K-1) , 2 (K-1) –1] and K is set to max(16, bit depth + 4), and the bit depth corresponds to the sample bit depth of the current block.

2. The video data prediction method according to claim 1, wherein: The refined predictor is pruned to K bits, where K is a positive integer.

3. The video data prediction method according to claim 2, wherein: K is set to max(14, bit depth+2) and the bit depth corresponds to the current block sample bit depth.

4. The video data prediction method according to claim 3, wherein: When the sample bit depth of the current block is equal to 10 bits, K is set to 14, and the second target range is equal to [-8192, 8191].

5. The video data prediction method according to claim 1, wherein: The pixel motion vector offset is derived based on second information regarding pixel position offset from the center of the target sub-block and the two or more control point motion vectors.

6. The video data prediction method according to claim 5, wherein: The prediction offset is derived based on a linear combination of the pixel motion vector offset in the x-direction and the pixel motion vector offset in the y-direction.

7. A video data prediction apparatus, wherein the video data prediction is performed by a video encoder or a video decoder using a set of codec tools including an affine mode, the apparatus comprising one or more electronic circuits or processors configured to: receiving input data related to a current block at a video encoder side, or receiving a video bitstream corresponding to compressed data including the current block at a video decoder side, wherein the affine mode is allowed for the current block; determining two or more control point motion vectors associated with the affine pattern for the current block; deriving a sub-block motion vector of a target sub-block of the current block from the two or more control point motion vectors of the current block; deriving a prediction offset of the target pixel of the target sub-block from the sub-block motion vector of the target pixel using first information including a pixel motion vector offset according to prediction refinement with optical flow; as well as encoding or decoding the target pixel of the target subblock using a modified predictor, wherein the modified predictor is generated by clipping the prediction offset to a first target range and combining the clipped prediction offset with an original predictor, or the modified predictor is generated by clipping a refined predictor to a second target range, and wherein the refined predictor is generated by combining the original predictor and the prediction offset, where the first target range corresponds to [–dILimit, dILimit–1], where dILimit is set to 1<<max(13, bit depth + 1), and the bit depth corresponds to the sample bit depth of the current block; the second target range corresponds to [–2 (K-1) , 2 (K-1) –1] and K is set to max(16, bit depth + 4), and the bit depth corresponds to the sample bit depth of the current block.

8. The video data prediction device according to claim 7, wherein: The refined predictor is pruned to K bits, where K is a positive integer.

9. The video data prediction device according to claim 8, wherein K is set to max(14, bit depth+2) and the bit depth corresponds to the current block sample bit depth.

10. The video data prediction device according to claim 9, wherein When the sample bit depth of the current block is equal to 10 bits, K is set to 14, and the second target range is equal to [-8192, 8191].

11. The video data prediction device according to claim 7, wherein: The pixel motion vector offset is derived based on second information regarding pixel position offset from the center of the target sub-block and the two or more control point motion vectors.

12. The video data prediction device according to claim 11, wherein The prediction offset is derived based on a linear combination of the pixel motion vector offset in the x-direction and the pixel motion vector offset in the y-direction.

Images