Despeckling of SAR images in contourlet domain using a new adaptive thresholding

Abstract

SAR is a powerful tool for producing high resolution remote sensing images under all weather conditions. But, during the backscattering process, the obtained images are contaminated with a special type of noise, called speckle. Speckle noise appears like a granular pattern in the SAR images and results in degrading the image quality. This work uses contourlet transform to effectively represent the directional information and intrinsic geometrical structures of the SAR image along the smooth contours (edges). Here, the speckled image is decomposed into directional sub-bands using contourlet transform followed by application of adaptive thresholding and finally, they are reconstructed to produce the denoised image. The adaptive thresholding uses a function with increasing and decreasing exponential functions to suppress the speckle content from the SAR image. The finally produced images prove their superiority in terms of speckle suppression and detail preservation.