A new SAR image despeckling using directional smoothing filter and method noise thresholding

Abstract

The removal of granular pattern multiplicative speckle noise is the major issue in Synthetic Aperture Radar (SAR) images. Theoretically, multiplicative noise is considered as the ratio of the standard deviation to the signal value. The proposed scheme works on db2 based 2D-Discrete Wavelet Transform (DWT) using wavelet thresholding and method noise. The scheme is designed to despeckle the simulated SAR images and real SAR images. It uses a hybrid combination of Directional Smoothing Filter (DSF), wavelet thresholding using enhanced Bayesian shrinkage rule and method noise for despeckling. After DWT decomposition, the approximate component is directed to DSF followed by method noise thresholding and detailed components are subjected to wavelet thresholding followed by method noise thresholding. The article justifies the efficient use of method noise and explains how its exact use can enhance the result of the algorithm over other efficient methods. The quality assessment of the proposed method is done by visual appearance and measures such as Peak Signal-to-Noise Ratio (PSNR), Structural similarity index metric (SSIM), Universal Image Quality Index (UIQI), Equivalent Numbers of Looks (ENL), Noise Variance (NV), Coefficient of Variation (CV), Mean Squared Error (MSE), Correlation Coefficient (CC) and preservation of mean values before and after despeckling. The effectiveness of the proposed method is demonstrated by matching it to well-known speckle noise removal methods on SAR images. The proposed method has the capability to be used in real time practical applications.