Edge generation image restoration method based on GAN network

Abstract

The invention discloses an edge generation image restoration method based on a GAN network. The method specifically comprises the following steps: firstly, collecting and arranging image data, makingan image data sample set, dividing the image data sample set into a training set and a test set, performing partial information shielding and graying processing on images of the training set to obtaingray images, and obtaining edge images and binary images of the images of the training set; establishing an edge generation image restoration model of the generative adversarial network, wherein theedge generation image restoration model of the generative adversarial network comprises two generators and two discriminators; and finally, inputting the image into the edge generation image restoration model, restoring the to-be-restored image, and outputting a restored clear image. The accuracy of the generated information is discriminated by using the discriminator for two times, the generatoris balanced, network weight and parameter sharing is realized during training, gradient descent, gradient disappearance and complex iterative operation are avoided, repair of damaged images is facilitated, and a good visual perception effect is generated.

Description

Technical field [0001] The invention belongs to the technical field of computer vision and image processing, and specifically relates to an edge generation image restoration method based on a GAN network. Background technique [0002] Visual sense organs are extremely important organs for humans to perceive external information. With the development of computer science and technology, images, as the main carrier of visual information acquisition and dissemination, have attracted more and more attention. A complete image is damaged or the content is missing, or the image resolution is relatively low, the image is blurred and other issues will cause people's visual perception to be incoherent. As a branch of image processing, image restoration technology has played an important role in solving the problem of image restoration. [0003] At present, traditional image restoration methods are divided into two categories. One is blind restoration based on digital image processing, mainl...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a method for image restoration based on GAN network edge generation, which solves the problems of incoherent repair of large-area missing images, unclear repair of texture edges, and low resolution in the prior art.

Images