Image coding and decoding and transmission method and system for neural network

Abstract

The invention discloses an image coding and decoding and transmission method and system for a neural network, and the method comprises the steps: S1, carrying out the down-sampling after the residual features of an image sequence are extracted; s2, a Gaussian filter is adopted to process the time and space pooling down-sampling matrix obtained in the step S1, and a smooth feature map is obtained; s3, optimizing down-sampling parameters to obtain an updated smooth feature map, and transmitting the updated smooth feature map; s4, at a receiving end, carrying out interpolation on the updated smooth feature map in the time direction, then carrying out interpolation in the height and width directions, and superposing the height and width interpolation feature map and the updated smooth feature map to obtain an up-sampling feature map; and S5, comparing the distribution difference between the updated smooth feature map and the up-sampling feature map, and optimizing an interpolation coefficient to obtain an optimal up-sampling feature map. According to the method, identifiable features and effective information of the image can be reserved, the coding and decoding efficiency is improved, and optimal image coding and decoding are realized.

Images