Planar mode improvement for intra prediction
Implementations of the disclosure provide a video processing method for performing intra prediction on a video block. The video processing method may include receiving reconstructed samples from a video frame in the bitstream including the video block. The video processing method may further include determining reference samples in neighboring video blocks of the video block according to a first slant line and a second slant line through a sample of the video block. The first slant line is along an interpolation direction and the second slant line is perpendicular to the first slant line. The video processing method may also include interpolating the sample of the video block based on the reference samples.
1. A video processing method for performing intra prediction on a video block comprising a plurality of samples from a bitstream, comprising:
receiving reconstructed samples from a video frame in the bitstream comprising the video block;
determining reference samples in neighboring video blocks of the video block according to a first slant line and a second slant line through a sample of the video block, wherein the first slant line is along an interpolation direction and the second slant line is perpendicular to the first slant line; and
interpolating the sample of the video block based on the reference samples.
2. The video processing method of claim 1 , wherein the interpolation direction is a block-level interpolation direction used for interpolating each of the plurality of samples in the video block.
3. The video processing method of claim 2 , further comprising:
determining the block-level interpolation direction for the video block.
4. The video processing method of claim 3 , wherein determining the block-level interpolation direction for the video block comprises:
calculating a rate-distortion cost of the video block when the video block is interpolated according to each of a plurality of candidate interpolation directions; and
determining the candidate interpolation direction associated with the rate-distortion cost having a minimum value as the block-level interpolation direction; and
signaling the block-level interpolation direction.
5. The video processing method of claim 3 , wherein determining the block-level interpolation direction for the video block comprises:
inferring a plurality of hypothesis interpolation directions for the video block based on neighboring reconstructed samples of the video block;
calculating a cost of the video block when the video block is interpolated according to each of a plurality of hypothesis interpolation directions; and
determining the hypothesis interpolation direction associated with the cost having a minimum value as the block-level interpolation direction.
6. The video processing method of claim 5 , wherein calculating the cost of the video block when the video block is interpolated according to a hypothesis interpolation direction comprises:
identifying neighboring reconstructed samples for each sample in the video block along the hypothesis interpolation direction and a direction perpendicular to the hypothesis interpolation direction;
calculating second order gradients of the neighboring reconstructed samples;
calculating a cost of each sample in the video block based on the second order gradients of the sample; and
calculating the cost of the video block as a sum of the calculated costs of the plurality of samples in the video block.
7. The video processing method of claim 6 , wherein calculating the second order gradients of the neighboring reconstructed samples comprises:
calculating first order gradients of the neighboring reconstructed samples;
normalizing the first order gradients based on distances between the neighboring reconstructed samples used for calculating the respective first order gradients; and
calculating the second order gradients based on the normalized first order gradients.
8. The video processing method of claim 1 , wherein the interpolation direction is a sample-specific interpolation direction, wherein a different sample-specific interpolation direction is used for interpolating each of the plurality of samples in the video block.
9. The video processing method of claim 8 , further comprising:
inferring a plurality of hypothesis interpolation directions for a sample of the video block based on neighboring reconstructed samples of the video block;
calculating a cost when the sample is interpolated according to each of the plurality of hypothesis interpolation directions; and
determining the hypothesis interpolation direction associated with the cost having a minimum value as the sample-specific interpolation direction.
10. The video processing method of claim 9 , wherein calculating the cost when the sample is interpolated according to a hypothesis interpolation direction comprises:
identifying neighboring reconstructed samples for the sample along the hypothesis interpolation direction and a direction perpendicular to the hypothesis interpolation direction;
calculating second order gradients of the neighboring reconstructed samples; and
calculating the cost of the sample based on the second order gradients of the sample.
11. The video processing method of claim 10 , wherein calculating the second order gradients of the neighboring reconstructed samples comprises:
calculating first order gradients of the neighboring reconstructed samples;
normalizing the first order gradients based on distances between the neighboring reconstructed samples used for calculating the respective first order gradients; and
calculating the second order gradients based on the normalized first order gradients.
12. The video processing method of claim 1 , wherein interpolating the sample of the video block based on the reference samples comprises:
generating fractional reference samples based on the reconstructed samples when reconstructed samples do not lie on the first slant line or the second slant line.
13. The video processing method of claim 12 , wherein each fractional reference sample is generated by interpolating two adjacent reconstructed samples nearest to the first slant line or the second slant line.
14. The video processing method of claim 1 , further comprising:
receiving a bitstream comprising a coded video block and an index signaling the interpolation direction.
15. A video processing apparatus for performing intra prediction on a video block comprising a plurality of samples from a bitstream, comprising:
a memory coupled to one or more processors; and
the one or more processors configured to:
receive reconstructed samples from a video frame in the bitstream comprising the video block;
determine reference samples in neighboring video blocks of the video block according to a first slant line and a second slant line through a sample of the video block, wherein the first slant line is along an interpolation direction and the second slant line is perpendicular to the first slant line; and
interpolate the sample of the video block based on the reference samples.
16. The video processing apparatus of claim 15 , wherein the interpolation direction is a block-level interpolation direction used for interpolating each of the plurality of samples in the video block.
17. The video processing apparatus of claim 16 , wherein one or more processors are further configured to:
determine the block-level interpolation direction for the video block.
18. The video processing apparatus of claim 15 , wherein the interpolation direction is a sample-specific interpolation direction, wherein a different sample-specific interpolation direction is used for interpolating each of the plurality of samples in the video block.
19. The video processing apparatus of claim 18 , wherein the one or more processors are further configured to:
infer a plurality of hypothesis interpolation directions for a sample of the video block based on neighboring reconstructed samples of the video block;
calculate a cost when the sample is interpolated according to each of the plurality of hypothesis interpolation directions; and
determine the hypothesis interpolation direction associated with the cost having a minimum value as the sample-specific interpolation direction.
20. A non-transitory computer-readable storage medium having stored therein instructions and a bitstream to be processed, when the instructions executed by one or more processors, cause the one or more processors to perform a video processing method for performing intra prediction on a video block comprising a plurality of samples from the bitstream, comprising:
receiving reconstructed samples from a video frame in the bitstream comprising the video block;
determining reference samples in neighboring video blocks of the video block according to a first slant line and a second slant line through a sample of the video block, wherein the first slant line is along an interpolation direction and the second slant line is perpendicular to the first slant line; and
interpolating the sample of the video block based on the reference samples.