Fast light super-resolution reconstruction dense residual error network

Abstract

The invention relates to a fast light super-resolution reconstruction dense residual error network. Based on a dual-channel deep residual error network (FLSR) of a convolutional network, a deep channel is mainly used for learning high-frequency texture information of an image, and a shallow channel is used for learning low-frequency information of of the image. In order to accelerate the convergence speed of the network, a residual error connection mode is added in the structure, in addition, the image detail information of the front convolution layer can be directly transmitted to the rear convolution layer through the residual error connection mode, and reconstruction of images with better quality is facilitated; and a dense connection mode which contributes to weakening gradient disappearance and improving the model performance is also adopted in the structure. While image quality evaluation indexes such as peak signal-to-noise ratio (PSNR), structural similarity (SSIM) and information fidelity (IFC) standard measurement are improved, parameters and computational complexity of the model are reduced, the reconstruction speed of the image is increased, and the method can be applied to actual life.

Description

technical field [0001] The invention belongs to the field of image super-resolution reconstruction, and relates to the improvement and application of an image super-resolution reconstruction method, in particular to a fast and lightweight super-resolution reconstruction dense residual network. Background technique [0002] With the development of society and the rapid progress of science and technology, the types of information obtained by people are diversified, and the ways of obtaining information are also increasing. Among them, the proportion of information obtained through vision exceeds 70%, and images and videos are currently the mainstay of visual information. The main carrier, which makes image processing research more and more extensive, and image processing technology is becoming more and more popular. [0003] Image resolution refers to the number of pixels contained in an image per inch, indicating the degree of resolution of scene details. It can be divided i...

AI Technical Summary

Problems solved by technology

[0007] Although deep networks perform well, most deep networks still contain many shortcomings

At present, deepening or widening the network has become the design trend of the network. These methods require a lot of calculation and memory consumption, and it is difficult to directly apply them in practice. Secondly, most of the current methods based on convolutional neural networks (CNNS) use a single channel. Shallow or deep networks are used to achieve super-resolution reconstruction, and preprocessing is required, but preprocessing may introduce new noise, and shallow channels are easy to lose high-frequency information of pictures; while deep networks are slow to converge and easy to reconstruct. Gradient explosion / vanishing phenomenon occurs

Images