Modified Fuzzy-Anisotropic Gaussian Kernel and CRB in Denoising SAR Image

Abstract

Radar speckle noise is often modeled as multiplicative noise for such that higher the intensity higher the speckle noise. As a result, the brighter pixel values are having more noise. The presence of speckle not only complicates visual image interpretation but also the classification of automated image is difficult in corrupted SAR image. Therefore, speckle has to be reduced before analyzing the SAR image.Thus, speckle is the main problem (mingled) in Synthetic Aperture Radar (SAR) images. Speckle is existed due to constructive and destructive interference of coherent signal. In order to reduce it, we approach enhanced kernel based filter. Till there are so many techniques are developed to remove speckle content in SAR system. But no proper technique as been developed to remove speckle content completely. In our project MMSE based filter technique is used. We propose a new integrated Fuzzy Anisotropic Gaussian Kernel (FAGK) for denoising Synthetic Aperture Radar (SAR) Images. Here, texture information lies on principal orientation should be multiplied with fuzzy membership function through the anisotropic Gaussian kernel. It presents Cramer -Rao Bound (CRB) which can be estimated by taking ensemble of texture modeled covariance matrix for different denoising methods. Later, CRB can be found for an index of speckle suppression. Thus, developed filter gives good result in preservation of texture and in structure enhancement. It also presents evaluation of speckle suppression ability, where an index named SMPI (Speckle Suppression and Mean Preservation Index). It compares CRB for the evaluation of SMPI index with different denoising methods