Image denoising method based on multi-channel GAN

Abstract

The invention discloses an image denoising algorithm based on a multi-channel GAN, belongs to the field of graphic processing, and particularly relates to an image denoising method based on the multi-channel GAN. The method comprises the steps of: firstly, based on a Unet derivative network, introducing residual block jump connection to extract features and fuse pixel-level features so as to effectively reserve image detail information; constructing a composite loss function based on the MSE, the feature perception and the adversarial loss to iteratively adjust the network so as to enable thegenerator and the discriminator to reach Nash equilibrium, thereby removing the image noise to the maximum extent; and finally, using the arithmetic mean weight to fuse the three-channel output information to obtain a final denoised image, so that the effect is that the subjective visual features are significant, and the proposed algorithm has low time consumption and good image denoising performance under different noise conditions.

Description

Technical field[0001]The invention belongs to the field of graphics processing, and in particular relates to an image denoising method based on a multi-channel generating confrontation network.Background technique[0002]In recent years, with the rapid progress of image processing technology, it has gained continuous attention in application fields such as medical imaging, satellite remote sensing, and intelligent monitoring. High-quality images are a prerequisite to ensure subsequent effective processing. However, the images are inevitably contaminated by noise during the acquisition and transmission process, which affects the reliability of subsequent image classification and recognition tasks. Therefore, how to remove the noise as much as possible without destroying the original features of the image to restore the original image as much as possible is one of the current hot issues in the field of image processing.[0003]In response to this problem, many effective denoising algorith...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem of poor denoising performance of traditional denoising algorithms, the present invention proposes a multi-channel image denoising method based on generative confrontation network

Images