IMAGE PROCESSING METHODS AND EQUIPMENT USING NEURAL NETWORK-BASED VOLUME REDUCTION FILTERING

Abstract

This invention proposes an image processing method and image processing device to acquire input data for decompression filtering based on target decompression points of at least one line perpendicular to the boundary of the blocks and encoded information about the target decompression points of at least one line; to acquire the feature map of the input data by inputting the input data into a first neural network; to acquire the decompression filter coefficient by inputting the feature map into a second neural network; to acquire the decompression filter compensation value by inputting the feature map into a third neural network; to acquire the decompression filter intensity by inputting the input data into a fourth neural network; and to acquire the decompression-filtered pixels by performing decompression filtering on the target decompression points of at least one line using the decompression filter coefficient, decompression filter compensation value, and decompression filter intensity.