图像去噪的改进迭代非局部平均滤波方法

Improved iterative non-local means filter for image denoising

提出了一种新的迭代非局部平均滤波的图像去噪方法.权系数的计算依赖每次迭代更新得到的图像,同时对迭代更新后得到的图像进行加权平均.这样就避免了权系数的计算以及加权平均所用的图像的不一致所带来的图像边缘模糊以及对比度不清晰的现象.还证明了新的迭代方法满足极大极小原则.实验结果表明,该方法去噪的同时能较好地保持图像的边缘以及细小结构.

An iterative non-local means filter for image denoising is proposed.For the proposed method,the computation of weights depends on the updated image,and for each iteration the weighted averaging is performed over the updated image.This avoids the problems that the edge of the image is blurry and that the image is not clear,which are caused by the disagreement of the computation of the weight coefficients and the computation of the averaging.The max-min principle is also guaranteed for the proposed iterative non-local means filter.Numerical examples illustrate that the proposed method removes noise well while preserving image edges and fine details.