A Comparison of the Bandelet, Wavelet and Contourlet Transforms for Image Denoising

The bandelet transform take advantage of the geometrical regularity of the structure of an image and is appropriate for the analysis of edges and textures of the images. Denoising is one of the most interesting and widely investigated topics in image processing area. The main problem in denosing is the tradeoff between the noise suppression and oversmoothing of image details. In order to solve that problem, in this paper we exploit the geometrical advantages offered by the bandelet transform to solve the problem of image denoising. We present the results obtained with the bandelet transform for denoising process with additive white Gaussian noise and salt and pepper noise. A comparison is made with those results obtained with wavelets and contourlets. We show that bandelets can outperform the wavelets and contourlets in terms of subjective and objective measures.