High-speed Noise Reduction in SAR Images Using Sparse Representation

One of the destructive factors in denotation scene details from the Synthetic Aperture radar images is the presence of speckle noise in it. Different method of image denoising have been proposed using the Sparse Representation (SR) Technique, which, eliminates the noise of the image properly by preserving details. Because of the Computational Complexity of these methods and the large dimensions of SAR images, their use in SAR images is challenging. This paper presents a new method in SAR image denoising by SR that reduce run time with preservation of image quality. In this method, the denoising performs in two step. Firstly, the image is filtered using a simple denoising method, and the removed details is then retrieved using the SR technique and added to the filtered image. By retrieving details from the heterogeneous regions of the image and using random sampling matrix in reconstruction image, the processing volume and the required memory in using SR technique are reduced. Simulation results show that this method, with preservation of image quality, has an operating time of about 0.2 of other methods.