A New Adaptive Image Denoising Method Combining the Nonsubsampled Contourlet Transform and Total Variation

Abstract

This paper presents a new adaptive image denoising scheme by combining the nonsubsampled Contourlet transform (NSCT) and total variation model. The original image is first decomposed using NSCT .Then the mean squared error (MSE) is estimated based on Stein’s unbiased risk estimation(SURE). The noise of each decomposed subband is reduced using the linear adaptive threshold function, which can be constructed based on the MSE, producing the preliminary primary denoised image after reconstruction. Then the preliminary primary denoised image is further filtered using the total variation model, producing the final denoised image. Experiments show that the proposed scheme can remove the pseudo-Gibbs artifacts and image noise effectively. Besides, it outperforms the existing schemes in regard of both the peak-signal-to-noise-ratio (PSNR) and the edge preservation ability.