NSCT De-Noising Algorithm Based on Image Partition and Noise Variance Estimation

Abstract

Aiming at the low de-noising effect of texture detail in the image, a NSCT de-noising method based on image partition and noise variance estimation was proposed. Since NSCT has the features of translation invariance and multi-directional selectivity, and noise has different impacts on different textures of image, the image can be divided into several blocks with the same size, then the noise variance and threshold of each block will be calculated, furthermore, NSCT is used to denoise each block, and finally, the blocks are merged. The experiments proved that by comparison with traditional NSCT de-noising algorithm, the proposed algorithm effectively preserved the texture detail information of original image.