Image super-resolution implementation algorithm based on information filtering network

Abstract

The invention discloses an image super-resolution implementation algorithm based on an information filtering network. The algorithm comprises a feature extraction module, a plurality of superposed information filtering modules and a reconstruction module, wherein the feature extraction module extracts features from a low-resolution image firstly, then superposes a plurality of information filtering modules, extracts residual information step by step, and finally, the reconstruction module aggregates the obtained high-resolution residual feature representation to generate a residual image; andelement-level item-by-item addition operation is performed on the residual image and the up-sampled low-resolution image. According to the image super-resolution implementation algorithm based on theinformation filtering network, due to the fact that the number of filters of each layer is relatively small, and group convolution is used, the network algorithm has the advantage of being high in execution speed.

Description

technical field [0001] The invention relates to an image super-resolution realization technology, in particular to an efficient image super-resolution realization algorithm based on an information filtering network. Background technique [0002] At present, image-based super-resolution has been widely studied, and the current mainstream methods are based on neural networks and example-based methods. [0003] Self-example-based methods exploit the self-similarity property to only extract example pairs from low-resolution images of different scales. This approach usually works well in images containing repetitive patterns and textures, but lacks the richness of semantic information external to the input image to produce satisfactory predictions for other classes of images. External example-based methods learn the mapping between low-resolution and high-resolution pairs from an external dataset. Such methods usually focus on how to learn a dense dictionary or popular space to...

AI Technical Summary

Problems solved by technology

But as a result, these methods require a lot of computational cost and memory consumption, which is not suitable for practical applications, such as in mobile terminals and embedded vision applications

Images