如何去除自然图像中的高斯白噪声是图像处理中的一个经典问题。基于小波收缩的NeighShrink降噪方法取得了很好的降噪效果，但是NeighShrink在所有小波子带上均使用了次优的universal阈值以及固定的邻域窗口尺寸，导致了较大的偏差，而且使得算法不健壮。为此，运用Stein的无偏风险估计改进了NeighShrink方法。该方法能够为每个小波子带确定最优的阈值和邻域窗口尺寸。实验结果显示，该方法取得了比NeighShrink更低的均方误差，也优于当前尖端的图像降噪算法—FeatShrink，其平均MSE大约低6%。

The denoising of a natural image corrupted by Gaussian noise is a classical problem in image processing. Chen et al. have proposed an effective wavelet thresholding scheme using incorporating neighboring coefficients, namely NeighShrink. NeighShrink’s disadvantages are to use a suboptimal universal threshold and fixed neighboring window size for every subband. In this paper, we improve NeighShrink using Stein’s unbiased risk estimation(SURE). Our method can determine an optimal threshold and neighboring window size for every subband. Experimental results show that our proposed method is significantly better than NeighShrink in all test examples. It also outperforms Balster et al’s FeatShrink which is a recent sophisticated image denoising algorithm.