Dual tree complex wavelet transform based SAR image despeckling

Abstract

In this paper, dual tree complex wavelet transform (DTCWT) based speckle reduction algorithm is proposed for despeckling SAR images. Here considering the dependency of the wavelet coefficients across the same scale and also different scales. The DTCWT coefficients of log transformed speckle free signal are modelled with a multivariate Cauchy pdf and signal-dependent noise are modelled with a multivariate Gaussian pdf. Expression for the dispersion parameter in the multivariate Cauchy pdf is derived by calculating a suitable fractional moment using Mellin transformation. Under these hypotheses, a generalized expression for the MAP estimator for the despeckled wavelet coefficients can be derived. This proposed method considering mean and directional median values of the neighbourhood window for calculating the interscale and intrascale dependency vector of wavelet coefficients. Experimental results show that the proposed method performs better than the existing MCMAP method in terms of equivalent number of looks, peak signal-to-noise ratio, and mean structural similarity index metric.