Image Laplace Denoising Based on Sparse Representation

Image denoising using sparse and redundant representation has drawn a lot of research attentions. For the existing denoising algorithms, the additive noise is always assumed to follow the Gaussian distribution. But in many application cases, the noise is not Gaussian. In this paper, we address the image Laplace denoising problem, where the additive noise is Laplace. Thus, our model is proposed by adopting the Bayesian MAP estimation theory. We operate this model on image patches and show how to solve it with linear programming. Our experimental results have shown good performance of our new method both in terms of peak signal-to-noise ratio (PSNR) and visually.