Abstract: To improve quality of medical image and give help to clinical diagnosis, an adaptive Non-local Means(NLM) denoising algorithm based on gradient information is proposed. The improved algorithm employies the gradient direction information to adaptively rotate the local similarity window so that more matching pixels are found. Based on least square idea, relationship model between the optimum parameter for the gradient amplitude threshold and the noise standard variance is established. Therefore, the adaptive selection of filtering parameter is realized. Experimental results show that the algorithm attains superior denoising effect and is appropriate for post-processing of medical image.

Key words: gradient, Non-local Means(NLM) filtering, medical image, adaptive filtering

摘要： 为改善医学图像的质量以利于临床诊断，提出一种基于梯度信息的自适应非局部均值去噪算法。利用梯度方向信息实现对局部相似窗的自适应旋转，从而搜寻到更多的匹配像素点。基于最小二乘思想建立最佳阈值与噪声标准差的关系模型，实现滤波参数的自适应选择。实验结果表明，该算法的去噪效果较好，可用于医学图像的后处理阶段。

关键词: 梯度, 非局部均值滤波, 医学图像, 自适应滤波

CLC Number: 

• TN911.73