Blind separation of mixed sources with curvelet de-noising

This paper investigates the technique of Fast Discrete Curvelet Transform (FDCT) de-noising with the Independent Component Analysis (ICA) for the separation of signals from noisy mixtures. Two approaches are presented for this purpose. In the first approach, the signals are separated using the fast ICA algorithm, and then curvelet thresholding is used to de-noise the results. The second approach uses curvelet thresholding to denoise the mixtures, and then the fast ICA algorithm to separate the signals from these mixtures. The simulation results show a better performance for image de-noising followed by separation. The Signal-to-Noise Ratio (SNR), and Peak Signal-to-Noise Ratio (PSNR) are used as quality evaluation metrics with signals and images, respectively.