Research On The Improvement Of Pepper And Salt Noise Identification And Filtering Algorithm

In the process of digital images formation and transmission,digital images are often affected and interfered by various factors to produce pepper and salt noise,which seriously affects the accuracy and quality of digital images.The pepper and salt noise may directly affect the resolution and sharpness of digital images and change the original image structure or image edges,thus causing great interference to the basic information extraction and subsequent processing of digital images.Therefore,in the process of image information extraction preprocessing,in order to suppress the image pepper and salt noise and improve the image quality,the image must be denoised.Image denoising is also one of the most important research topics in the field of digital image information preprocessing.As an important research topic in the field of image processing,many researches on pepper and salt denoising at home and abroad have used identification and filtering algorithm to identify and filter the pepper and salt noise in the image,so as to restore the image.At present,as far as the identification algorithm is concerned,it still faces some problems,such as low noise identification accuracy and high false alarming ratio.At present,as far as the filtering algorithm is concerned,it still faces some problems,such as the filtered image is not clear and discontinuous,the mean square error and the peak value signal to noise ratio are not ideal.Therefore,in order to improve the image's pepper and salt noise identification accuracy and quality of filter,on the basis of analyzing and studying the advantages and disadvantages of common noise identification algorithms(neighborhood identification method,connected domain identification method)and filtering algorithms(mean filtering method,median filtering method,wiener filtering method and wavelet filtering method),two pepper and salt noise identification algorithms and one salt and pepper noise filter algorithm have been designed.The two pepper and salt noise identification algorithms are BPNN_NI noise identification algorithm(BP Neural Network Noise Identification)and FM_N noise identification algorithm(Fuzzy Math Neighborhood),which are used to solve the problem of accurately identifying pepper and salt noise.The pepper and salt noise filter algorithm is GAK filtering algorithm(Genetic Algorithm Kalman),which is used to solve the quality problem of pepper and salt noise filtering.In view of the fact that there are many picture data sets and the data sets are available for the training of neural network by BPNN_NI identification algorithm,the BPNN_NI identification algorithm based on BP neural network is designed to accurately identify pepper and salt noise by researching and analyzing BP neural network and combining the advantages of neighborhood information identification method and connected domain information identification method.The input value and network structure of BPNN_NI identification algorithm are designed using neighborhood pixel median value,neighborhood pixel mean value,connected domain pixel median value and ROAD pixel value.According to the principle of gradient descent,the training method and decision criterion of BPNN_NI identification algorithm have been designed.Finally,Matlab software is used to identify the noise points in the image with pepper and salt noise.The simulation results show that the BPNN_NI identification algorithm has low noise leakage numbers and false alarming ratio of the pepper and salt noise points in the identification image,and has good identification effect on the pepper and salt noise.In view of the fact that there is no large data set for the training of neural network by BPNN_NI identification algorithm,the FM_N identification algorithm has been designed to accurately identify pepper and salt noise by researching and analyzing the fuzzy mathematics and combining with the neighborhood identification method.8 neighborhood identification method,24 neighborhood identification method and ROAD pixel value are used as the input value of the FM_N identification algorithm.Then the fuzzy set and membership function,evaluation index and fuzzy rules are set to form the fuzzy discriminator.At last,the identification criterion are added to the fuzzy discriminator and the identification result of pepper and salt noise are outputted.The simulation results show that the FM_N identification algorithm has low noise leakage numbers and false alarming ratio,a good identification effect of pepper and salt noise,and the FM_N identification algorithm has no dependence on the size of image data set.The new image filtering algorithm GAK is designed by researching and analyzing the initial state value selection of the extended kalman filter and genetic algorithm.By using genetic algorithm,the kalman filter gain is optimized and the the extended kalman filter is optimized to filter out pepper and salt noise efficiently.Firstly,the principle of extended kalman filter and genetic algorithm is analyzed.Then the principle of GAK filtering algorithm is proposed,the function and core programming of GAK filtering algorithm are described in detail,and the solving steps of GAK filtering algorithm are designed.Finally,Matlab software is used to adjust the parameter values of GAK filtering algorithmand the image denoising simulation of pepper and salt noise is carried out.The simulation results show that the image restored by GAK filtering algorithm has high resolution and GAK filtering algorithm has better applicability and suppression ability for processing salt and pepper noise.