An Adaptive Threshold Method Based on the Local Energy of NSCT Coefficients for Image Denoising

In this paper, we propose a new adaptive denoising method based on nonsubsampled Contour let transform(NSCT). Traditional Wavelet provides only three directional components so that its geometrical property is not well, and Contour let transform lacks translation invariance, therefore NSCT is developed. The proposed algorithm can adapt different thresholds on different scales and different directions. Further more, we use different thresholds in a directional subband according to local energy of NSCT coefficients to overcome the disadvantages of the unified threshold de-noising method and other fixed thresholds, which cause the image fuzzy distortion because of "over-killed". The experimental results prove that the algorithm outperforms existing schemes in both peak-signal-to-noise-ratio (PSNR) and visual quality.