现代谱估计与K-S检验相结合的正则参数选择  

Choice of Regularization Parameter by Modern Spectral Estimation Combined with K-S Test

作　　者：罗晓华[1,2] 

机构地区：[1]重庆交通大学图书馆,重庆400074 [2]重庆大学电气工程学院,重庆400044

出　　处：《东莞理工学院学报》2010年第5期54-58,共5页Journal of Dongguan University of Technology

摘　　要：将残差的NCP同K-S(Kolmogorov-Smirnov)判据相结合来确定正则参数,并以图象恢复为例进行仿真。首先,将残差的NCP同理论值进行比较,求出最佳正则参数。而后,以图象恢复为例进行了讨论,结果表明,用现代谱估计的Burg法得到的恢复图象比传统谱估计FFT的图象质量和稳定性要好。In this paper,the regularization parameter is found by NCP of the residual vector combined with K-S test,and as example,image resumption is imitated.At first,NCP of the residual vector is compared with the theoretical value,and found optimal regularization parameter.Secondly,as example,image resumtion is discussed.The result shows that the quality and an stabilities of the image by the modern spectral estimation（Burg method） is better more than the classical spectral estimation（FFT）.

关 键 词：病态问题 正则化方法 图象恢复 K-S检验 NCP方法 

分 类 号：TN911.7[电子电信—通信与信息系统]