Image Denoising Based on Second Generation Bandelets and Multi-level Thresholding

As an updated multiresolution geometry analysis tool, second generation bandelets can make full use of intrinsic geometry regularity of images, and then provide an optimal representation of the underlying high-dimensional functions adaptively. In this paper, a novel approach to image denoising is presented, which is based on second generation bandelets and combined with multi-level thresholding. Numerical tests applied on natural images contaminated additive Gaussian white noise show that our method provides a significant improvement both in terms of image visual fidelity and in terms of peak signal-to-noise ratio. Comparisons are made with wavelet image noise-removal results