Method for determining specific linear model and video processing device

1. A method performed by a video decoding device for predicting and reconstructing a chroma block using a specific linear model, the method comprising:

receiving a bitstream;

determining a plurality of linear model parameters of each of a plurality of linear models;

determining locations of the chroma block and a luma block, that corresponds to the chroma block, from a video frame retrieved from the bitstream;

determining a first template region neighboring the chroma block and a second template region neighboring the luma block, wherein the first template region is reconstructed as a chroma template reconstruction and the second template region is reconstructed as a luma template reconstruction;

determining a chroma template prediction of the first template region for each of the plurality of linear models based on the luma template reconstruction and the plurality of linear model parameters of each linear model;

determining a template cost for each of the plurality of linear models based on a difference between the chroma template reconstruction and a corresponding chroma template prediction;

selecting the specific linear model from the plurality of linear models for predicting the chroma block based on the template cost of each of the plurality of linear models;

using the specific linear model to generate a first chroma prediction result of the chroma block;

determining, based on the bitstream, a chroma residual block corresponding to the chroma block; and

reconstructing the chroma block based on the first chroma prediction result and the chroma residual block.

2. The method according to claim 1 , wherein determining the first template region neighboring the chroma block comprises:

determining at least one neighbouring chroma region of the chroma block as the first template region, each of the at least one neighbouring chroma region comprising a plurality of reconstructed chroma samples.

3. The method according to claim 2 , wherein the at least one neighbouring chroma region of the chroma block comprises at least one of a first chroma region above the chroma block and a second chroma region on the left side of the chroma block.

4. The method according to claim 1 , wherein determining the second template region neighboring the luma block comprises:

determining at least one neighbouring luma region of the luma block, each of the at least one neighbouring luma region comprising a plurality of luma samples;

downsampling the at least one neighbouring luma region and the luma block to obtain a downsampled luma block and at least one downsampled neighbouring luma region of the downsampled luma block;

determining the at least one downsampled neighbouring luma region as the second template region corresponding to the luma block.

5. The method according to claim 4 , wherein downsampling the at least one neighbouring luma region and the luma block comprises:

downsampling the at least one neighbouring luma region and the luma block by using a downsample filter, wherein the downsample filter is determined based on a video format of a video frame corresponding to the chroma block.

6. The method according to claim 4 , wherein the at least one downsampled neighbouring luma region of the downsampled luma block comprises at least one of a first luma region above the downsampled luma block and a second luma region on the left side of the downsampled luma block.

7. The method according to claim 1 , wherein the plurality of linear models comprises a first linear model, and determining the chroma template prediction of the first template region for each of the plurality of linear models based on the luma template reconstruction and the plurality of linear model parameters of each linear model comprises:

determining a first linear function based on the plurality of linear model parameters of the first linear model;

determining the chroma template prediction corresponding to the first linear model by using the first linear function to process the luma template reconstruction,

wherein the plurality of linear model parameters of the first linear model is characterised by α and β, and a sample in position (i, j) of the chroma template prediction Pred template (i, j) corresponding to the first linear model is characterised by:

Pred template ( i,j )=α×rec′ L ( i,j )+β

wherein rec′ L (i, j) is a luma sample in position (i, j) of the luma template reconstruction.

8. The method according to claim 1 , wherein the plurality of linear model parameters of each of the plurality of linear models is generated based on a plurality of reference sample regions.

9. The method according to claim 8 , wherein the plurality of reference sample regions is different from each other and the plurality of reference sample regions comprises at least one neighbouring region of the chroma block.

10. The method according to claim 1 , wherein the plurality of linear models comprises a first linear model, and determining the template cost for each of the plurality of linear models based on the difference between the chroma template reconstruction and the corresponding chroma template prediction comprises:

determining a first difference between the chroma template reconstruction and the chroma template prediction corresponding to the first linear model as the template cost of the first linear model.

11. The method according to claim 1 , wherein the specific linear model has a lowest template cost among the plurality of linear models.

12. The method according to claim 1 , further comprising:

selecting a reference linear model from the plurality of linear models and using the reference linear model to generate a second chroma prediction result of the chroma block; and

combining the first chroma prediction result and the second chroma prediction result into a specific chroma prediction result of the chroma block.

13. The method according to claim 12 , wherein the chroma block is designed with a combining flag, which indicates whether to combine the first chroma prediction result and the second chroma prediction result into the specific chroma prediction result of the chroma block.

14. The method according to claim 12 , wherein combining the first chroma prediction result and the second chroma prediction result into the specific chroma prediction result of the chroma block comprises:

determining the specific chroma prediction result of the chroma block via performing a weighted linear combination on the first chroma prediction result and the second chroma prediction result.

15. The method according to claim 14 , wherein weightings corresponding to the first chroma prediction result and the second chroma prediction result are determined based on the template costs of the specific linear model and the reference linear model, respectively.

16. The method according to claim 1 , further comprising:

determining, from the bitstream, a specific chroma mode flag associated with the chroma block,

wherein the specific chroma mode flag indicates whether to use the specific linear model as a prediction mode to generate the first chroma prediction result of the chroma block.

17. The method according to claim 16 , further comprising:

in response to the specific chroma mode flag being determined to have a first value, forgo decoding other chroma mode flags corresponding to the plurality of linear models; and

in response to the specific chroma mode flag being determined to have a second value, decoding the other chroma mode flags corresponding to the plurality of linear models.

18. The method according to claim 1 , wherein the video frame comprises a chroma component having the chroma block.

19. A video decoding device, comprising:

a non-transitory computer-readable medium storing one or more computer executable instructions; and

a processor coupled to the non-transitory computer-readable medium and configured to execute the one or more computer-executable instructions to:

receive a bitstream;

determine a plurality of linear model parameters of each of a plurality of linear models;

determine locations of a chroma block and a luma block, that corresponds to the chroma block from a video frame retrieved from the bitstream;

determine a first template region neighboring the chroma block and a second template region neighboring the luma block, wherein the first template region is reconstructed as a chroma template reconstruction and the second template region is reconstructed as a luma template reconstruction;

determine a chroma template prediction of the first template region for each of the plurality of linear models based on the luma template reconstruction and the plurality of linear model parameters of each linear model;

determine a template cost for each of the plurality of linear models based on a difference between the chroma template reconstruction and a corresponding chroma template prediction;

select a specific linear model from the plurality of linear models for predicting the chroma block based on the template cost of each of the plurality of linear models;

use the specific linear model to generate a first chroma prediction result of the chroma block;

determine, based on the bitstream, a chroma residual block corresponding to the chroma block; and

reconstruct the chroma block based on the first chroma prediction result and the chroma residual block.