Iterative dual tree wavelet transform with posterior probability for sar despeckling

Speckle is a natural distortion produced by the interaction of backscattered waves from the target in active coherent imaging sensors like synthetic aperture radar (SAR). The presence of speckle in SAE images mainly restricts its usage in terms of ground target detection, information retrieval, and scene investigation. Numerous filtering techniques have been available in the literature to reduce speckles. Recently, wavelet transform approaches can be employed for the denoising purposes, which make use of a threshold value on the noisy image for sparse wavelet representation. However, the noisy image coefficients surpassing the threshold are the cause of spurious noise spikes over the discontinuities. Therefore, this study develops an iterative dual tree wavelet transform with posterior probability for SAR despeckling process. The proposed method works on the wavelet domain for efficient speckle noise reduction. The speckled image is fed to the Dual tree complex wavelet transform to produce six oriented wavelets which are then employed to the thresholding method. For fixing the threshold value, the posterior probability is used in each subband which in turn gives the threshold value concerning the distribution of coefficients. Furthermore, thresholding is applied on oriented wavelets for despeckling process. Entropy is calculated in each iteration for detecting the number of decomposition levels. The performance validation of the proposed model take place in terms of different measures. The experimental results stated that the proposed system is highly accurate than conventional 2D-DTCWT model.