基于同步噪声选择非局部平均除噪声的最优参数  

Selection of the optimal parameter in non-local means denoising based on synchronous noise

作　　者：刘鹏[1] 刘定生[1] 李国庆[1] 刘志文[1] 

机构地区：[1]中国科学院对地观测与数字地球科学中心,北京100086

出　　处：《光电子．激光》2011年第7期1107-1111,共5页Journal of Optoelectronics·Laser

基　　金：中国科学院对地观测与数字地球科学中心主任基金;国家自然科学基金共同资助项目(41001265)

摘　　要：提出基于同步噪声和信噪比(SNR)自动选择非局部平均(NCM)的核函数参数的算法。为了正确估计平滑图像中的剩余噪声,一幅纯噪声图像作为同步噪声与观测图像同步进行计算,同步噪声作为图像剩余噪声的估计值。为了保持同步噪声与剩余噪声的统计特性相似,提出了一种新的NCM算子,依据图像的特征对同步噪声进行平滑,能够保持两种噪声有相似的统计特性。基于同步噪声推导剩余噪声、图像SNR和NCM的核函数参数间的关系式,提出了基于此关系式优化选择模型参数的算法流程。对不同噪声水平的不同图像进行了充分的实验,实验结果验证了该方法的有效性。In this paper a new method for selecting the optimal kernel width for non-local means denoising is proposed.The rule of the stopping time is based on signal noise ratio（SNR）.In order to correctly estimate the level of noise in non-local means denoising,a pure synthesis noise as an auxiliary is synchronously smoothed with the image.The auxiliary noise is taken as the estimation of the remaining noise in the estimated image.A new non-local means smoothing operator that can ensure the synchronous changing of the synthesis noise and the real noise is proposed in this paper.The relationship between image detail and the variance of the synchronous noise is established,and a rule for the optimal kernel width of non-local means denoising is deduced.Experiments confirmed the effectiveness of the proposed method.

关 键 词：非局部平均(NCM) 同步噪声 最优参数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]