Nonsubsampled Shearlet-based image denoising using multiscale products

In this paper, a new denoising algorithm based on translation-invariant nonsubsampled Shearlet transform (NSST) using multiscale products threshold is proposed. After analyzing the dependence of the NSST coefficients across scales, space, and directions, the adaptive threshold is then applied to the multiscale products of the NSST coefficients not directly applied to the NSST coefficients. The approach introduced here presents two major advantages: (a) NSST gets more directional subbands which capture the anisotropic information of natural image; (b) The multiplicating operation enhances the significant features while weakening noise. Experimental results show that the proposed scheme outperform other state-of-art denoising methods.