NSCT-based image denoising using 3-D directional windows

Nonsubsampled Contourlet Transform (NSCT) is an effective representation tool in image denoising for its outstanding asymptotic rate and shift-invariance property. Proposed is a novel NSCT-based image denoising algorithm makes use of directional dependence. We describe the NSCT coefficients into 4D models, and select threshold after filtering with 3D directional windows. Experiment results show that the algorithm can highly shrinks the noise coefficients, enhances the edge coefficients and obtains higher peak signal-to-noise ratio.