Depth residual convolution neural network image denoising method based on PReLU

Abstract

The invention relates to a depth residual convolution neural network image denoising method based on PReLU, based on deep convolution neural network, combined with Gaussian noise simulating unknown real noise image denoising task, in this paper, a deep convolution neural network for image denoising is proposed, which uses PReLU activation function instead of Sigmoid and ReLU function, increases residual learning and reduces mapping complexity, and adopts optimized network training techniques and network parameter settings to improve the denoising ability of the network. Compared with other existing denoising algorithms, the present invention performs very well under various Gaussian noise environments in which the standard variance is mixed, and the detailed information in the image can bewell preserved while the noise is eliminated.

Description

technical field [0001] The present invention relates to the technical field of image denoising, in particular to a PReLU-based deep residual convolutional neural network image denoising method. Background technique [0002] In the process of image acquisition or transmission, people will generate noise due to some irresistible factors, such as the influence of external conditions such as light, temperature, weather and image imaging equipment, and the influence of internal conditions such as resistance, electromagnetic and component interference. Noisy images with degraded quality and blurred features will affect subsequent information dissemination and image analysis and processing. Therefore, in the field of image processing, image denoising technology has always been an indispensable research topic and a necessary process in the preprocessing part of many related image algorithms, which has very important theoretical value and practical significance. [0003] According t...

AI Technical Summary

Problems solved by technology

[0004] For the spatial domain denoising algorithm, the noise in the image can be effectively weakened by this algorithm, but it also brings another trouble - the blurring of the image and the loss of details

The principle of spatial domain denoising algorithms is to process the whole image without considering the details of the image, and the problems of loss of details and blurred images have not been improved; the switch median filter algorithm has great defects. The filtering effect of the image with low noise is very good, but if the image contains low noise, its denoising performance drops sharply

Second, how to set the threshold will have an extremely serious impact on the filtering results

However, when the image noise intensity increases, the useful information that can be used inside the image decreases, and the denoising effect of the BM3D method based on the information inside the noise image will become worse.

Images