Image denoising with an optimal threshold and neighbouring window

NeighShrink is an efficient image denoising algorithm based on the decimated wavelet transform (DWT). Its disadvantage is to use a suboptimal universal threshold and identical neighbouring window size in all wavelet subbands. In this paper, an improved method is given, which can determine an optimal threshold and neighbouring window size for every subband by the Stein’s unbiased risk estimate (SURE). Its denoising performance is considerably superior to NeighShrink and also outperforms SURE-LET, which is an up-to-date denoising algorithm based on the SURE. It is well known that increasing the redundancy of wavelet transforms can significantly improve the denoising performances. The proposed method is also extended to the redundant dual-tree complex wavelet transform (DT-CWT). Experiments demonstrate that the proposed method on the DT-CWT achieves better results than some of the best denoising algorithms published currently.