Multi-stage denoising system and method for color image

Abstract

The invention provides a multi-stage denoising system and method for a color image. The multi-stage denoising system comprises a channel separation module, an intra-channel denoising network and an inter-channel denoising network, wherein the channel separation module is used for separating the color image with noise into original noise images of a plurality of channels; the intra-channel denoising network is used for carrying out preliminary denoising on an original noise image; and the inter-channel de-noising network is used for further de-noising according to the preliminary de-noising result and the original noise image to obtain a clean image. In specific implementation, a denoising task is divided into two stages: denoising in an image channel is carried out by utilizing spatial dimension information in a single channel, and then inter-channel denoising is further carried out by utilizing associated information between channels; meanwhile, when denoising processing is carried out, an error feedback mechanism is added to an up-sampling part and a down-sampling part, and the denoising effect is enhanced.

Description

technical field [0001] The invention relates to the technical field of computer vision image processing, in particular to a multi-stage denoising system and method for color images. Background technique [0002] In the process of image acquisition, transmission and storage, it is inevitable to be disturbed by various noises. Image noise blurs the image, and sometimes even conceals image features, and at the same time adversely affects image visual effects and subsequent data analysis. Therefore, how to efficiently denoise real images and improve image quality is an important research topic in the field of computer vision. [0003] At present, traditional image denoising algorithms include PCA denoising, DCT denoising, BM3D denoising, etc. Among them, BM3D is a traditional denoising method with better effect at present. It obtains the block evaluation value by searching for similar blocks and filtering in the transform domain. , and finally weight each point in the image to...

AI Technical Summary

Problems solved by technology

However, none of the above-mentioned deep learning methods have taken into account that: simultaneous extraction of information within and between image channels will increase the difficulty of denoising; compared with a single feed-forward method, the human visual system is closer to a feedback mechanism; input image information will Lost or cleaned as the network deepens

Therefore, the above-mentioned methods based on deep learning are relatively limited in image denoising.

Images