Neural network based filter in video coding
In one implementation, to perform in-loop filtering of a version of reconstructed samples of a block, only a single offset parameter is signaled in the bitstream. Based on the version of reconstructed samples, a pixel-wise weight mask is generated using a neural network. Because the neural network parameters are known at both the encoder and decoder, these parameters need not to be signaled in the bitstream. The single offset parameter scaled by the weighted mask is used to adjust the samples in the block. Thus, even though only a single offset parameter is used, the samples are adjusted by pixel-wise offsets. The neural network may also take other parameters, such as quantization parameters and picture types as input. Further, there can be multiple neural networks that generate different weight masks, where different offsets are signaled and one or more of the neural networks are to be selected for filtering.
1. A method, comprising:
accessing a version of reconstructed samples of a region of a picture;
generating a weight for each sample of a plurality of samples in said region based on said version of reconstructed samples of said region, using a neural network;
encoding an offset parameter in a bitstream, wherein neural network parameters of said neural network are not transmitted in said bitstream; and
filtering said plurality of samples in said region, wherein a sample of said plurality of samples in said region is adjusted based on a weight for said sample and said offset parameter for said region, and wherein said offset parameter for said region is scaled by said weight for said sample and said scaled offset is added to said sample to adjust said sample.
2. The method of claim 1 , wherein a plurality of neural networks are used, and wherein each network of said plurality of neural networks corresponds to a classification of said region.
3. The method of claim 1 , further comprising:
generating another weight for each sample of said plurality of samples in said region based on said version of reconstructed samples of said region, using another neural network; and
encoding another offset parameter corresponding to said another neural network for said region in said bitstream, wherein neural network parameters for said another neural network are not encoded in said bitstream,
wherein said sample in said region is adjusted further based on said another weight for said sample and said another offset parameter for said region, and wherein said another offset parameter for said region is scaled by said another weight for said sample and said scaled another offset is further added to said sample to adjust said sample.
4. The method of claim 3 , wherein a sum of said scaled offset and said scaled another offset is added to said sample in said region to filter said sample.
5. An apparatus, comprising one or more processors, wherein said one or more processors are configured to:
access a version of reconstructed samples of a region of a picture;
generate a weight for each sample of a plurality of samples in said region based on said version of reconstructed samples of said region, using a neural network;
encode an offset parameter for said region in a bitstream, wherein neural network parameters of said neural network are not transmitted in said bitstream; and
filter said plurality of samples in said region, wherein a sample of said plurality of samples in said region is adjusted based on a weight for said sample and said offset parameter for said region, and wherein said offset parameter for said region is scaled by said weight for said sample and said scaled offset is added to said sample to adjust said sample.
6. The apparatus of claim 5 , wherein a plurality of neural networks are used, and wherein each network of said plurality of neural networks corresponds to a classification of said region.
7. The apparatus of claim 5 , wherein said one or more processors are further configured to:
generate another weight for each sample of said plurality of samples in said region based on said version of reconstructed samples of said region, using another neural network; and
encode another offset parameter corresponding to said another neural network for said region in said bitstream, wherein neural network parameters for said another neural network are not encoded in said bitstream,
wherein said sample in said region is adjusted further based on said another weight for said sample and said another offset parameter for said region, and wherein said another offset parameter for said region is scaled by said another weight for said sample and said scaled another offset is further added to said sample to adjust said sample.
8. The apparatus of claim 7 , wherein a sum of said scaled offset and said scaled another offset is added to said sample in said region to filter said sample.
9. A method, comprising:
accessing a version of reconstructed samples of a region of a picture;
generating a weight for each sample of a plurality of samples in said region based on said version of reconstructed samples of said region, using a neural network;
decoding an offset parameter corresponding to said neural network for said region from a bitstream, wherein neural network parameters of said neural network are not transmitted in said bitstream; and
filtering said plurality of samples in said region, wherein a sample of said plurality of samples in said region is adjusted based on a weight for said sample and said offset parameter for said region, and wherein said offset parameter for said region is scaled by said weight for said sample and said scaled offset is added to said sample to adjust said sample.
10. The method of claim 9 , wherein a plurality of neural networks are used, and wherein each network of said plurality of neural networks corresponds to a classification of said region.
11. The method of claim 9 , further comprising:
generating another weight for each sample of said plurality of samples in said region based on said version of reconstructed samples of said region, using another neural network; and
decoding another offset parameter corresponding to said another neural network for said region in said bitstream, wherein neural network parameters for said another neural network are not encoded in said bitstream,
wherein said sample in said region is adjusted further based on said another weight for said sample and said another offset parameter for said region, and wherein said another offset parameter for said region is scaled by said another weight for said sample and said scaled another offset is further added to said sample to adjust said sample.
12. The method of claim 11 , wherein a sum of said scaled offset and said scaled another offset is added to said sample in said region to filter said sample.
13. An apparatus, comprising one or more processors, wherein said one or more processors are configured to:
access a version of reconstructed samples of a region of a picture;
generate a weight for a respective sample of a plurality of samples in said region based on said version of reconstructed samples of said region, using a neural network;
decode an offset parameter for said region from a bitstream, wherein neural network parameters of said neural network are not transmitted in said bitstream; and
filter said plurality of samples in said region, wherein a sample of said plurality of samples in said region is adjusted based on a weight for said sample and said offset parameter for said region, and wherein said offset parameter for said region is scaled by said weight for said sample and said scaled offset is added to said sample to adjust said sample.
14. The apparatus of claim 13 , wherein a plurality of neural networks are used, and wherein each network of said plurality of neural networks corresponds to a classification of said region.
15. The apparatus of claim 13 , wherein said one or more processors are further configured to:
generate another weight for each sample of said plurality of samples in said region based on said version of reconstructed samples of said region, using another neural network; and
decode another offset parameter corresponding to said another neural network for said region in said bitstream, wherein neural network parameters for said another neural network are not encoded in said bitstream,
wherein said sample in said region is adjusted further based on said another weight for said sample and said another offset parameter for said region, and wherein said another offset parameter for said region is scaled by said another weight for said sample and said scaled another offset is further added to said sample to adjust said sample.
16. The apparatus of claim 15 , wherein a sum of said scaled offset and said scaled another offset is added to said sample in said region to filter said sample.