Image Restoration Based on Improved Generative Adversarial Networks

Abstract

Image restoration aims to recover image information from broken or missing regions, which plays an important role in the field of computer vision. There has been a major breakthrough in image restoration by using deep neural network models, but it is not effective in restoring above high-resolution image details. In order to solve this problem, this paper proposes a generative adversarial network with embedded channels and spatial attention, simultaneously, makes the network obtain a larger perceptual field by increasing the size of the expanded convolution, which leads to induce the network to learn the image details and broken edges better, so as to improve the restoration effect on details. The network proposed in this paper is trained and learned on a public datasets, and it is shown in the experimental results that the effect of the method on image restoration is improved to a certain extent.