Improved regularization based blind image deconvolution using PSO algorithm for PMMW images application

Blind deconvolution algorithms in image processing estimate the original image and the true Point Spread Function (PSF) simultaneously with prior information about the PSF or the original image. This is an ill-posed problem and requires regularization to be solved. In addition to the type of regularization functions, the value of regularization parameters can drastically affect the output result. In this paper, we propose a Particle Swarm Optimization (PSO) algorithm for selecting optimum values of regularization parameters in blind image deconvolution. The algorithm has been tested on standard images and then compared with non-improved ones, in terms of three standard metrics. A real Passive Millimeter Wave (PMMW) image blurred by an unknown PSF is also used in this algorithm to obtain a sharp deblurred image with an estimate of the PSF. Simulation results show that the proposed method improves the quality of the estimated PSF and the deblurred image.