A Novel Inter-Scale Correlation Image Denoising Method Based on Dual-Tree M-Band Wavelet

A novel inter-scale correlation image denoising method based on Dual-tree M-band wavelet (DTT) is proposed in this paper. Dual-tree M-band wavelet transform is a shift-invariant, multi-scale and multi-direction transform based on a Hilbert pair of wavelets initially proposed by N. Kingsbury. Improving upon Xu's denosing algorithm based on wavelet inter-scale correlation, a new correlation modeling is provided between each high frequency detail subimage and corresponding M subimages in adjacent lower frequency scale. In the new algorithm, signal and noise are distinguished by the strength of the correlation, and combined with threshold functions. The experiment result shows that comparing with the classical denoising methods, for example, wavelet denoising method, Dual-tree complex wavelet denoising method, contourlet denoising method and so on..., the proposed denoising method achieves an excellent balance between suppressing noise effectively and preserving as many image details and edges as possible.