IMAGE DENOISING THROUGHMULTI-SCALE LEARNT DICTIONARIES

Over the last decade, a number of algorithms have shown promising results in removing additive white Gaussian noise from natural images, and though different, they all share in common a patch based strategy by locally denoising overlap-ping patches. While this lowers the complexity of the prob-lem, it also causes noticeable artifacts when dealing with large smooth areas. In this paper we present a patch-based denois-ing algorithm relying on a sparsity-inspired model (K-SVD), which uses a multi-scale analysis framework. This allows us to overcome some of the disadvantages of the popular algo-rithms. We look for a sparse representation under an already sparsifying wavelet transform by adaptively training a dictio-nary on the different decomposition bands of the noisy image itself, leading to a multi-scale version of the K-SVD algo-rithm. We then combine the single scale and multi-scale ap-proaches by merging both outputs by weighted joint sparse coding of the images. Our experiments on natural images in-dicate that our method is competitive with state of the art al-gorithms in terms of PSNR while giving superior results with respect to visual quality. Index Terms — K-SVD, sparse, dictionary, multiscale, wavelet, denoising