Multiplicative noise removing method based on deep neural network

Abstract

The invention provides a multiplicative noise removing method based on a deep neural network, and mainly solves the problems that the peak signal-to-noise ratio is low and the speed is low when imagemultiplicative noise is removed in the prior art. The method is characterized in that a nonlinear mapping relation between a noisy image and a grayscale image is fitted through a trained deep neural network, and is used for removing the multiplicative noise. The method comprises the steps of acquiring a grayscale image set (X) and a noise-containing image set (Y), constructing a deep neural network model; using the grayscale image set (X) and a noise-containing image set (Y) to train the weight and bias of each layer of the deep neural network model, and inputting a to-be-denoised image into atrained deep neural network model. The output result of the deep neural network model is a denoised grayscale image. According to the method, the peak signal-to-noise ratio of the denoised image is improved, and the speed of removing the multiplicative noise is increased. The method can be applied to the occasions of for image denoising and preprocessing including image classification, object recognition, edge detection and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to an image multiplicative noise removal method, in particular to a deep neural network-based multiplicative noise removal method, which can be applied to image classification, target recognition, edge detection, etc. In the case of noise preprocessing. Background technique [0002] 80% of the information obtained by humans from the external world comes from vision. With the digitization of images and the popularization of image equipment, as well as the rapid development of computer technology, digital images have become the main information carrier and the source of information in many scientific fields. However, in the process of image acquisition, noise is introduced due to various reasons, which has a certain impact on image quality and visual effects. According to the interference relationship between noise and image signal, noise can be divided into additive noise and...

AI Technical Summary

Problems solved by technology

Although this method improves the speed of multiplicative noise removal, due to the use of block matching operations in this method, the denoising speed is still limited and cannot achieve real-time effects.

[0006] The image denoising method based on the neural network can better fit the nonlinear mapping from the noisy image to the denoised image, but the shallow neural network is not comprehensive enough to learn the characteristics of the image, and cannot learn more detailed structures in the image. information, which eventually leads to a low peak signal-to-noise ratio of the denoised image

Images