Method and device for picture encoding and decoding

1. A method for video decoding, comprising:

accessing a first reconstructed version of an image block from video data;

in-loop filtering the first reconstructed version of the image block by applying an inference process based on an in-loop filter implemented by a first neural network to form a second reconstructed version of the image block adapted to be used as reference for prediction; and

out-of-loop filtering a version based on the second reconstructed version of the image block by applying an inference process based on an out-of-loop filter implemented by a second neural network to form a third reconstructed version of the image block adapted for display,

wherein the in-loop filter and the out-of-loop filter are trained jointly using a joint loss function.

2. The method of claim 1 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.

3. The method of claim 1 , wherein the version based on the second version is the second version or an intermediate representation of the output of the in-loop filter corresponding to an output of a layer before a last layer of the in-loop filter.

4. A non-transitory information storage medium storing program code instructions for implementing the method according to claim 1 .

5. A method for video encoding, comprising:

accessing a first reconstructed version of an image block of a picture from video data; and

in-loop filtering the first reconstructed version of the image block using an in-loop filter implemented in the form of an inference process of a first neural network to form a second reconstructed version of the image block adapted to be used as reference for prediction,

wherein the in-loop filter is trained jointly with an out-of-loop filter implemented in the form of an inference process of a second neural network used for out-of-loop filtering a version based on the second reconstructed version of the image block to form a third reconstructed version of the image block adapted for display, the in-loop filter and the out-of-loop filter being trained using a joint loss function.

6. The method of claim 5 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.

7. A non-transitory information storage medium storing program code instructions for implementing the method according to claim 5 .

8. The method of claim 5 , wherein the joint loss function comprises a first term representative of a first loss function between the image block and the second reconstructed version of the image block.

9. The method of claim 8 , wherein the first loss function is a L1 or L2 norm.

10. The method of claim 5 , wherein the joint loss function comprises a second term representative of a second loss function between the image block and the third reconstructed version of the image block.

11. The method of claim 10 , wherein the second loss function is determined using a Generative Adversarial Network.

12. The method of claim 10 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.

13. An apparatus for video decoding comprising one or more processors configured to perform:

accessing a first reconstructed version of an image block from video data;

in-loop filtering the first reconstructed version of the image block by applying an inference process based on an in-loop filter implemented by a first neural network to form a second reconstructed version of the image block adapted to be used as reference for prediction; and

out-of-loop filtering a version based on the second reconstructed version of the image block by applying an inference process based on an out-of-loop filter implemented by a second neural network to form a third reconstructed version of the image block adapted to display,

wherein the in-loop filter and the out-of-loop filter are trained jointly using a joint loss function.

14. The apparatus of claim 13 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.

15. The apparatus of claim 13 , wherein the version based on the second version is the second version or an intermediate representation of the output of the in-loop filter corresponding to an output of a layer before a last layer of the in-loop filter.

16. An apparatus for video encoding comprising one or more processors configured to:

accessing a first reconstructed version of an image block of a picture from video data; and

in-loop filtering the first reconstructed version of the image block using an in-loop filter implemented in the form of an inference process of a first neural network to form a second reconstructed version of the image block to be used as reference for prediction,

wherein the in-loop filter is trained jointly with an out-of-loop filter implemented in the form of an inference process of a second neural network used for out-of-loop filtering a version based on the second reconstructed version of the image block to form a third reconstructed version of the image block to be displayed, the in-loop filter and the out-of-loop filter being trained using a joint loss function.

17. The apparatus of claim 16 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.

18. The apparatus of claim 16 , wherein the joint loss function comprises a first term representative of a first loss function between the image block and the second reconstructed version of the image block.

19. The apparatus of claim 16 , wherein the joint loss function comprises a second term representative of a second loss function between the image block and the third reconstructed version of the image block.

20. The apparatus of claim 19 , wherein the first reconstructed version of the image block is used as an additional input to train the out-of-loop filter.