Evaluation of Hard Thresholding of Curvelet Coefficients for Speckle Reduction in SAR images

Due to the damaging effects of Speckle noise in accessing the information on radar images, reduction of these effects has been considered by many researchers. In this study, speckle reduction of radar images has been discussed based on curvelet transform. This paper describes an adaptive method of threshold estimation based on curvelet transform for removing speckle noise from Synthetic Aperture Radar (SAR) images. The estimation of the threshold value is carried out by analyzing the statistical parameters of the curvelet subband coefficients like arithmetic mean and geometric mean. In this algorithm first multiplicative speckle noise is transformed into an additive one by taking the logarithm of the original speckled image. Second curvelet transform is taken of logarithmically transformed image. Then based upon the statistical parameters of the curvelet coefficients of subbands, threshold values are found out. This threshold value is used in hard thresholding technique to remove the noisy curvelet coefficients. Then the inverse transform is applied to get the denoised image. Evaluation parameters like edge preservation factor, Equivalent Number of Looks, Mean Square Error and so on has been used for evaluating the performance of the proposed technique quantitatively. The results have been compared then to those obtained by other widely-used adaptive filters including Frost, Gamma, Kuan, Lee and hard thresholding wavelet based filters. The results of this comparison show that the proposed method offers better results than the above-mentioned filters (Mean Square Error, Normalized Mean Square Error and Mean Absolute Error indices are reduced 32%,32% and 5% respectively). In addition, the curvelet based filter is capable of preserving edges too.