Image restoration method based on adversarial generative network

Abstract

The invention relates to an image restoration method based on an adversarial generative network, and the method comprises the steps: S1, constructing an image restoration training network: adding a generated SE-RestNet network into a generative network, adding a discrimination SE-RestNet network into a discrimination network, and obtaining an image restoration training network; S2, conducting training: extracting a plurality of original images from a training data set; performing mask processing to obtain a plurality of training images, using the generated SE-RestNet network in the generated network to generate a training restoration image, and then using the discrimination SE-RestNet network in the discrimination network to discriminate whether the restored image is true or false, and after the discrimination network reaches a balance state, using the trained generation network as an image restoration network. According to the image restoration method based on the generative adversarial network provided by the invention, an SE-RestNet network block is added into the generative adversarial network, so that the restored image ensures the image structure and semantic coherence, the image restoration effect is better, and no restoration trace exists.

Description

Technical field [0001] The invention relates to the field of image modification, in particular to an image restoration method based on a confrontation generation network. Background technique [0002] Image restoration is an image processing technology that uses existing information in the image to repair missing information in the image or remove specific information in the image under the premise of ensuring the clarity of the image and its semantic continuity. The core challenge of this technology is to synthesize visually realistic and semantically reasonable pixels for the missing areas so as to be consistent with existing pixels. Image restoration has important practical significance, especially in the protection of works of art, restoration of old photos, image-based rendering and computer photography. [0003] Traditional image restoration methods mostly use image-level features to deal with restoration problems. Patch-Match method proposes to find the best matching patch...

AI Technical Summary

Problems solved by technology

[0011] There are still deficiencies in the existing image modification technology, which needs to be improved and improved.

Images