基于改进LOG算子的非局域均值图像降噪

摘要

In order to improve the non-local means image denoising algorithm,we use a improved laplacian of gaussian operator to substitute the exponential weight function.Because the improved laplacian of gaussian operator can preserve image edge better and can be obtained by the derivative of gaussian weight function.Furthermore,basing on the feature of local similarity,we truncate the weight function.Experimental results on several images show that our improved algorithm gives better performance.Otherwise,the computation time can be reduced because of symmetry of weighted Euclidean distance simultaneity and avoiding repeat distance calculating of two pixels.%为了改进非局域均值图像去噪的性能,利用二阶导数可以更好地保护图像边缘的特性,通过高斯权值函数获得高斯-拉普拉斯算子,并对其进行修正代替指数权值函数。再根据局域像素块相似性特点将权值函数进行截断。在多个典型图像上的实验结果表明,提出的改进方法获得了较好的降噪性能。另外,在利用欧氏距离对称性的同时,避免了两像素点间距离的重复计算,降低了算法的处理时间。