Lapped transform domain SAR image despeckling using neighboring coefficients

In this paper, a SAR image despeckling algorithm based on Lapped transform(LT) is proposed. The motivation of using LT is that they are robust to oversmoothing and can better preserve the textural information present in the images. The dyadic rearranged lapped orthogonal transform (LOT) coefficient, of the logarithmically transformed reflectance, is modified using a shrinkage rule which uses the sum of squares of all the LOT coefficients within a squared neighborhood and Donoho´s universal threshold. It is assumed that, a large dyadic rearranged LOT coefficient in a subband will usually have large coefficients in its neighbourhood. The proposed algorithm employs a scaling factor for the Donoho´s universal threshold to get better despeckling performance. Experimental results on real SAR images show that the proposed method outperforms several well known despeckling methods and can achieve both the speckle reduction and texture preservation simultaneously with much less computational complexity.