The Research Of Blind Image Separation Based On Independent Component Analysis And Wavelet Threshold Denoising

Blind Source Separation(BSS),an active research field in modern signal processing fields,refers to recovering source signals from the observations received by sensors with the source signals and transmit channel unknown.Blind Source Separation technology has made great achievements in application like image and speech processing,biomedical signal processing and so on.In the field of image processing,the theory and technology of blind source separation has gone deep into the separation of images,image denoising,and the feature extraction of image and other aspects of image processing.In order to obtain the closer combination of the BSS technology with the image processing,a new fast independent component analysis(Fast-ICA)algorithm is proposed for BSS,investigate the performance of the comparison of the traditional Fast-ICA algorithm and the improved Fast-ICA algorithm in the noisy and the noisy-free blind separation processing.In the following,a method based on the improved Fast-ICA and the wavelet transform(WT)is used to accomplish noisy blind image separation,which also improves the performance of BSS.Firstly,analysis the fundamental theory of BSS and the main method to solve BSS-independent component analysis(ICA),In various ICA methods,the Fast-ICA is an excellent algorithm and widely used in many fields which is based on the fixed points.Fast-ICA finds the de-mixing matrix that optimizes the nonlinear contrast function by the quadratic convergence of Newton iteration method.To accelerate the convergence speed of the algorithm,an improved Fast-ICA algorithm with eight-order convergence of Newton iterative method is proposed.Secondly,in order to compare the effectiveness and feasibility of the improved Fast-ICA algorithm,the basic Fast-ICA algorithm and the improved Fast-ICA algorithm are successively applied to the noise-free ICA model.The simulation results show that the improved Fast-ICA has fewer iterations and faster convergence rate for the corresponding of the traditional Fast-ICA and the Fast-ICA with fifth-order convergence of Newton iteration method in the nearly same separation performance.Lastly,noise cannot be ignored in the practical engineering applications.A method that based on wavelet transform and the improved Fast-ICA algorithm is used to accomplish the separation of noisy BSS.The improved Fast-ICA algorithm is used in the separation and the methods are used in the denoising option such as wavelet transform,the mean filtering and the median filtering.Finally,the denoising and separation performance are analyzed by the means of a series of evaluation index.The simulation results show that the method of the combination of improved Fast-ICA and wavelet transform can be used to realize the better performance in the noisy BSS for multiple images separation.