Image Denoising in Shearlet Domain by Adaptive Thresholding
Zhe Chen · Journal of Information and Computational Science · 2013
Shearlet transform contains some useful and many good properties in signal analysis, such as sparsity, multi-scale and multi-direction. Compared to the existing multi-scale geometry analysis, for instance, the Curvelet and Contourlet, Shearlets are theoretically optimal in representing images and particularly have the ability to fully capture directional and geometrical features. Furthermore, it can be easily implemented. Therefore, it is very suitable for the image denoising. A new adaptive thresholding algorithm in Shearlet domain which based on SURE-LET is proposed in this paper. Noisy image is firstly transformed into Shearlet domain; and then a pointwise thresholding is performed on the Shearlet coefficients based on SURE-LET in order to remove additive white Gaussian noise; at last, the image is reconstructed by the inverse Shearlet transform. Experimental results show that the performance of the proposed method is comparable to the state-of-the-art denoising algorithms, such as NLM and BM3D, but with the advantage of its low complexity and better computational efficiency.