Sonar image denoising using adaptive processing of local patches

Sonar images are highly affected by signal-dependent multiplicative speckle noise. Denoising is required in sonar images to distinguish a number of different regions by analyzing the image. In this paper, we propose sonar image denoising based on a signal independent additive Gaussian noise model. The sparse representation of the sonar images is exploited in the denoising method. The noisy image, image patches and blocks of patches are denoised using Principal Component Analysis and Singular Value Decomposition methods. Comparison of different methods is done using different non reference image performance evaluation criteria.