Mixed Noise Filtering Algorithm For Color Images Based On BM3D

Image is one of the most important ways for human beings to perceive the external world.Compared with hearing and touch,image has the largest information density and can contain more abundant information.Therefore,most of human information is derived from images.With the development of information science and technology,the demand of computer intelligent processing of information is increasing.However,image degradation and noise are inevitable in the process of transmission and conversion.This greatly interferes with the understanding of images.It can be seen that the significance of image denoising is not only to improve the quality of the image,but also to play an important role in subsequent image recognition and image understanding.So in the field of digital image processing,image noise filtering has always been one of the most important and basic research topics.Image noise can be divided into many kinds according to its mathematical characteristics.This paper mainly focuses on the mixed noise filtering algorithm composed of Gauss noise and impulse noise.Mixed noise filtering becomes very complex because of the superposition of two kinds of noise,but in practice,this kind of situation is prevalent.To solve this problem,the following works are done in this paper:In this study,our goal is to remove the mixed noise composed of additive white gaussian noise(AWGN)and random value impulse noise(RVIN).First of all,we analyze the characteristics of color image mixed noise,understand its structure through mathematical modeling of color image mixed noise,and provide ideas for the design of denoising algorithm.Secondly,the DWM algorithm is improved.In order to detect and remove impulse noise more accurately,a recursive function is used to adjust the threshold value.As a result,the self-adaptability of the algorithm is significantly improved and the application range is expanded.The experimental results show that the improved algorithm has stronger denoising performance.Finally,the improved directional weighted median filtering algorithm is combined with block matching 3d filtering algorithm(BM3D),and a new color image hybrid noise filtering algorithm is proposed.Specifically,the mixed noise filtering is divided into two stages.In the first stage,the improved directional weighted median filter algorithm is used to preliminarily process the polluted image and remove most of the impulse noise.In the second stage,BM3 D algorithm is used to re-process the contaminated image to remove the remaining noise components.This algorithm is suitable not only for mixed noise but also for single pulse noise or gaussian noise.It is especially suitable for some cases where the image noise type is not known.The experimental results show that the algorithm has better performance than the existing algorithms.