Edge detection in speckled SAR images using SWT and multiscale product

In this paper, stationary wavelet transform (SWT)-based despeckling and edge detection algorithm is proposed for SAR images. The first part of this paper describing a Despeckling algorithm based on maximum a posteriori probability (MAP) criterion. The MAP solution is based on the assumption that wavelet coefficients have a known distribution. The wavelet coefficients of the speckle free image and the signal dependent speckle noise are modelled with Laplacian distribution and Gaussian distribution respectively. Then a closed form solution of the MAP estimation is developed. The performance of this despeckling method can be improved by using a segmented approach, where each wavelet subband is divided into different classes of heterogeneity according to the texture energy of the wavelet coefficients of noise-free reflectivity. The second part of this paper describing an edge detection algorithm based on a combination of wavelet coefficients at different scales. Here, first calculating the pointwise maximum across horizontal, vertical and diagonal subbands. Then these three subband maximum values are combined through pointwise multiplication. The performance of this method is tested on various images.