Image super-resolution method based on hierarchical residual neural network

Abstract

The invention discloses an image super-resolution method based on a hierarchical residual neural network. A hierarchical residual learning strategy and a double-domain enhancement module are adopted to construct a hierarchical residual neural network model, the hierarchical residual neural network model is used for learning a complex mapping relationship between a low-resolution image and a high-resolution image, and then the trained network model is utilized to reconstruct the high-resolution image from the input low-resolution image, wherein the hierarchical residual neural network model iscomposed of a feature extraction layer, a feature mapping layer and a feature fusion layer, the feature extraction layer is composed of a convolution module, the feature mapping layer is composed of aplurality of double-domain enhancement modules in a cascaded mode, and the feature fusion layer is composed of an up-sampling module and a convolution module. According to the method, a better high-resolution image can be reconstructed, and the method has the advantages of being small in network model parameter quantity and high in calculation efficiency.

Description

technical field [0001] The invention belongs to the technical fields of image processing and machine vision, and in particular relates to an image super-resolution method based on a hierarchical residual neural network. Background technique [0002] The resolution of images collected by existing digital cameras, video cameras and other imaging devices is often low, which seriously affects subsequent image analysis and understanding. [0003] Traditional image super-resolution methods use manual feature modeling, which is difficult to describe realistic and complex super-resolution problems, which limits its performance and application range. [0004] Existing image super-resolution methods based on deep learning improve performance by increasing network depth and sharing parameters, but the network model is too large to be used in actual scenarios. While lightweight super-resolution methods perform poorly. These factors restrict the application of existing image super-reso...

AI Technical Summary

Problems solved by technology

Although these methods achieve good results, they lack efficient network models, and their performance still cannot meet the needs of practical applications.

Images