Speckle Removal from SAR Images in the Lapped Transform Domain Using Adaptive Threshold Based on Despeckling Evaluation Indexes

We propose in this paper an efficient and adaptive threshold estimation method for suppressing speckle in synthetic aperture radar (SAR) images, based on Lapped transform (LT). Lapped orthogonal transform (LOT) is robust to oversmoothing, can preserve texture components better and has less computational complexity than orthogonal discrete wavelet transform (DWT). Since LOT is block transform, the dyadic remapping of transform coefficients are carried out first and then the subband coefficients are thresholded. We estimate the ‘optimal’ threshold value using two despeckling performance indexes namely equivalent number of looks (ENL) and edge saving index (ESI). The threshold value is then used in soft thresholding function to shrink the noisy LOT coefficients. Due to the use of despeckling performance indexes in threshold estimation, the test on real SAR images show that the proposed method notably suppresses speckle with good edge preservation and achieves better performance than other related methods.