Block-wise content-adaptive online training in neural image compression

An apparatus for video decoding, comprising:

• processing circuitry configured to
• decode, from a coded bitstream, neural network update information for a first neural network component in a video decoder and for a second neural network component that is downstream of the first neural network component in the video decoder such that an input of the second neural network component is based on an output of the first neural network component, the first neural network component and the second neural network component being configured with pretrained parameters, the first neural network update information corresponding to a first block in an image to be reconstructed and indicating the first neural network component, a first layer of the first neural network component, and a first replacement parameter to replace a pretrained parameter in the first layer of the first neural network component, and corresponding to a second block in the image to be reconstructed and indicating the second neural network component, a second layer of the first neural network component, and a second replacement parameter to replace a pretrained parameter in the second layer of the second neural network component;
• update the first layer of the first neural network component in the video decoder based on the first replacement parameter and decode the first block based on the updated first layer of the first neural network component and based on the pretrained parameters of the second neural network component;
• and
• update the second layer of the second neural network component in the video decoder based on the second replacement parameter and decode the second block based on the updated first layer of the first neural network component and the updated second layer of the second neural network component.

A non-transitory computer-readable storage medium storing a program executable by at least one processor to perform:

• decoding, from a coded bitstream, neural network update information for a first neural network component in a video decoder and for a second neural network component that is downstream of the first neural network component in the video decoder such that an input of the second neural network component is based on an output of the first neural network component, the first neural network component and the second neural network component being configured with pretrained parameters, the neural network update information corresponding to a first block in an image to be reconstructed and indicating the first neural network component, a first layer of the first neural network component, and a first replacement parameter to replace a pretrained parameter in the first layer of the first neural network component, and corresponding to a second block in the image to be reconstructed and indicating the second neural network component, a second layer of the first neural network component, and a second replacement parameter to replace a pretrained parameter in the second layer of the second neural network component;
• updating the first layer of the first neural network component in the video decoder based on the first replacement parameter and decoding the first block based on the updated first layer of the first neural network component and based on the pretrained parameters of the second neural network component;
• and
• updating the second layer of the second neural network component in the video decoder based on the second replacement parameter and decoding the second block based on the updated first layer of the first neural network component and the updated second layer of the second neural network component.