Comparative Study of Transform Domain Filters with Modified Particle Swarm Optimization for Speckle Noise Suppression of SAR Images

This paper gives a comparison of transform domain filters such as wavelet transform, contourlet transform, Bandelet transform and curvelet transform with Modified PSO (MPSO) for the feature enhancement, edge preservation and despeckling of SAR images. In this paper, despeckling and preservation of edges are combined with improved gain function which optimize the curvelet coefficients for the enhancement of quality of denoised image. Further, global search algorithm such as MPSO algorithm is applied for the best result. Increase in convergence speed and reduction in premature convergence are possible with MPSO which produces new learning scheme and a mutation operator. Curvelet transform with MPSO is compared to wavelet, contourlet and bandelet transform with MPSO. Experimental results show that the curvelet transform with MPSO gives better quality of image but time complexity is high and other transforms have low time complexity.