Image denoising method and system based on multi-scale expansion convolution residual network

Abstract

The invention discloses an image denoising method and a system based on a multi-scale expansion convolution residual network, and the method comprises the steps: obtaining a training data set, and carrying out the cutting of an image in the training data set, and obtaining a block image; constructing a network model, adopting a mode of combining batch normalization and residual learning, adoptingan optimal mixed expansion rate mode, and introducing a multi-scale structure to obtain an end-to-end image denoising model; setting hyper-parameters of a network model, and selecting a loss functionand an optimization method to train the image denoising model to obtain a trained image denoising model; and converting the noise picture, inputting the converted noise picture into a trained image denoising model, performing average operation on the obtained picture, and outputting the denoised picture. According to the method, the denoising performance of the network can be ensured while the parameter quantity of the network is reduced, and the sharp edge and the fine detail information of the picture can be reserved while the noise of the picture is removed.

Description

technical field [0001] The invention relates to the technical field of image denoising, in particular to an image denoising method and system based on a multi-scale dilated convolution residual network. Background technique [0002] Image denoising is a classic and active problem in computer vision. In the process of acquiring images, due to the interference of irresistible external environmental noise such as light, temperature, and weather, and the influence of components such as resistance and electromagnetics on imaging equipment, the image will generate noise during the digitization and imaging process, thereby affecting Image quality, which in turn affects the later image dissemination and image processing, such as action recognition, image segmentation, etc. Therefore, image denoising technology has very important research significance. [0003] According to the characteristics of different images and the law of noise, at present, image denoising algorithms are main...

AI Technical Summary

Problems solved by technology

At present, the most popular methods include DnCNN, Ffdnet, etc. Although these methods have achieved good denoising effects, these methods usually improve the denoising performance at the expense of the number of parameters, which brings difficulties to practical applications.

Images