Real image denoising method based on multi-scale selection feedback network

Abstract

The invention discloses a real image denoising method based on a multi-scale selection feedback network, comprising: constructing a multi-scale selection module MSB for extracting multiple receptive field scale features; constructing a multi-scale selection feedback network MSFB, including a shallow feature extraction unit , multiple concatenated MSBs, image reconstruction units, and degradation models; construct two dual tasks for image denoising: predict noise-free images from original noisy images, and degrade from predicted noise-free images to noisy images; exploit MSFB at multiple times Two dual tasks are repeatedly performed in the step, and multi-level iteration is performed; in the iteration, the high-level semantic information output by the deep MSB of the previous time step is selectively fed back to the input of the shallow MSB of the next time step, and the MSFB is trained iteratively; In the training process, the optimization goal is to minimize the dual loss, and the peak signal-to-noise ratio is used as the evaluation index of the network performance; the noisy image is input into the trained MSFB for denoising, and the denoised image is output.

Description

technical field [0001] The invention relates to the fields of computer vision and image processing, in particular to a real image denoising method based on a multi-scale selection feedback network. Background technique [0002] When the real image is processed, stored, and transmitted in the acquisition system, a variety of complex noises will be generated, resulting in the loss of structural details and the degradation of image quality. And image noise will also be subject to such decomposition and synthesis. The electrical system and external influences in these processes complicate the precise analysis of image noise. Most of the denoising methods that exist today are based on known synthetic additive white Gaussian noise, but tend to perform poorly in real-world noisy images. [0003] Image denoising is a typical image restoration task. The characteristic of direct image-to-image conversion results in an infinite number of correspondences for noisy images in the clean ...

AI Technical Summary

Problems solved by technology

[0006] In view of this, the present invention proposes a real image denoising method based on a multi-scale selective feedback network, adding additional supervision in the noise domain to the network, which not only reduces the excessive dependence on clean and high-quality training data, but also makes the network sensitive to noise changes. It is more robust to solve the problems of poor denoising effect and high complexity of the denoising model in current denoising methods for real noisy images

Images