Analysis of Blind Separation of Noisy Mixed Images Based on Wavelet Thresholding and Independent Component Analysis

This paper investigates the technique of wavelet threshold de-noising with Independent Component Analysis (ICA) for noisy image separation. In the first approach, noisy mixed images are separated using fast ICA algorithm and then wavelet thresholding is used to de-noise. The second approach uses wavelet threshold to de-noise and then use the fast ICA algorithm to separate the de-noised images. The simulation results show better performance of image separation followed by denoising rather than the other way round. Peak Signal to Noise Ratio (PSNR), Improved Signal to Noise Ratio (ISNR), Signal to Noise Ratio (SNR) and Root Mean Square Error (RMSE) are used to evaluate quality of separated images. Amari error and structural similarity index (SSIM) is computed for the separation quality measurement. noiseless data and these algorithms perform poorly in the presence of noise (14). In this paper, noisy multiple channel blind signal separation algorithms based on wavelet thresholding are investigated. In the first approach, noisy mixed images are separated using fast ICA algorithm and then soft wavelet thresholding is used to de-noise. Second approach uses soft wavelet thresholding to de-noise and then the use of the fast ICA algorithm to separate the de-noised images.