基于梯度信息的快速非局部均值图像去噪算法  

Fast Non-local Means Image Denoising Algorithm Based on Gradient Information

作　　者：肖鹏[1] 刘平平[1] 陈幼平[1] 

机构地区：[1]华中科技大学机械科学与工程学院,湖北武汉430074

出　　处：《机械与电子》2010年第11期3-6,共4页Machinery & Electronics

基　　金：国家自然科学基金资助项目(50905064);教育部留学回国人员科研启动基金资助项目(2008)

摘　　要：非局部均值图像去噪算法具有优秀的去噪效果,但是算法复杂度高,不能应用于高速图像处理系统中。为提高算法执行速度,使其拥有更广泛的应用,提出了基于图像梯度信息的快速非局部均值图像去噪算法。该算法把原始图像划分为大梯度区域和小梯度区域。利用非局部均值算法对大梯度区域去噪,以保证图像边缘的清晰度;利用局部加权平均算法对小梯度区域去噪,以保证灰度变化不大的区域信息的完整性和准确性。算法能提高非局部均值滤波速度,而且能够有效保存图像边缘和细节。Non-local means algorithm can achieve a state-of-the-art denoising result at the cost of a high complexity,which is not adaptable enough to response a high-speed image processing.In order to lower the complexity,rise the algortithm speed as well as broaden its application,this paper proposed a fast non-local means denoising algorithm based on image gradient.Divide the original image into a large gradient region and a small one.By respectively using non-local means denoising algorithm in large gradient region and local weighted average filtering in small gradient region,we can both ensure the clarity of image edge and a higher credibility of the pixel in the neighborhood so that the completeness and veracity of small gradient region remains unaffected.This proposed algorithm can improve the speed of non-local means filter significantly and reserve the image edges and details effectively as well.

关 键 词：图像去噪 非局部均值 梯度 SOBEL算子 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]