Research On Image Recognition And Filtering Algorithm Of Salt And Pepper Noise

With the rapid development of science and technology,digital image processing technology has been inseparable from people’s production and life.Digital image processing usually includes image acquisition,image storage,image display,data communication,image analysis and other links.In these links,digital images are often affected and interfered by various factors,such as external electromagnetic interference in the process of image transmission,internal timing errors of image sensors,image encoding and decoding frame loss,etc.These accidents easily lead to salt-and-pepper noise in images,which seriously affects the visual quality of digital images and greatly interferes with the subsequent image information extraction and processing.Therefore,image denoising has become one of the important research topics in digital image preprocessing.In order to improve the overall denoising performance of salt and pepper noise images,this thesis compares and analyzes the advantages and disadvantages of common noise recognition algorithms.Two kinds of recognition algorithms,BP＿NR(Back Propagation Noise Recognition)and FC＿NR(Fuzzy Control Noise Recognition),are designed to realize effective recognition of salt and pepper noise.Based on the comparison and analysis of the advantages and disadvantages of common filtering algorithms,the Adaptive Variable Window Filtering(AV＿WF)algorithm is designed to effectively remove salt and pepper noise points.BP＿NR algorithm is a noise recognition algorithm based on BP neural network,which aims to solve the problems of low noise recognition accuracy and high parameter misjudgment rate of traditional recognition algorithms.The traditional recognition algorithm only takes the neighborhood mean and neighborhood value as the feature vector of the noise image.In order to better protect the edge part of the image,this thesis introduces a new feature vector ROAD factor on this basis,then normalizes these feature vectors,sets network parameters,builds network structure,and trains the algorithm model according to the principle of gradient descent method.Finally,the effectiveness of BP＿NR algorithm on noise recognition was verified by Matlab software.The FC＿NR algorithm is a noise recognition algorithm based on a fuzzy controller.Compared with the BP＿NR algorithm,the recognition performance of the FC＿NR algorithm does not depend on the size of the training set,and the algorithm implementation is independent.On the basis of the FC＿NR algorithm,which is based on the traditional eight neighborhood recognition method and the twenty-four neighborhood recognition method,the FC＿NR algorithm uses the noise intensity function to further judge the fuzzy relations that may occur in the discriminant process,and sets the fuzzy set,the membership function,the evaluation index and the fuzzy rules to form a double fuzzy controller.Finally,experiments are conducted to verify the effectiveness of the FC＿NR algorithm for noise recognition.The AV＿WF algorithm is a filtering algorithm based on Standard Median Filtering(SMF),which aims to solve the problems of low peak signal-to-noise ratio and poor image detail protection of traditional filtering algorithms.The filter window is designed to fit the noise density in the region,and the noise points in the image are mainly processed.When restoring the gray value of the image,the weight of gray value of the original pixel and the estimated value of noise is considered at the same time,and the original gray value of the polluted pixel is recalculated by using the membership function of the pixel noise intensity,so that the restored image is closer to the image without noise pollution.