预估计最小平方误差初始值方法改善交替映射算法性能  

Improving the Performance of Alternating Projection Demosaicking Algorithm by Pre-Estimating Minimum Square Error Initialization

作　　者：单宝堂[1] 沈庭芝[1] 崔玉红[2] 赵鹏[1] 胡宇[1] 

机构地区：[1]北京理工大学信息科学技术学院电子工程系,北京100081 [2]装备指挥技术学院信息装备系,北京101416

出　　处：《北京理工大学学报》2007年第5期436-440,共5页Transactions of Beijing Institute of Technology

基　　金：香港理工大学研发项目

摘　　要：改进交替映射算法以提高反马赛克插值的峰值信噪比(PSNR)性能.采用预估计最小平方误差思想,改进交替映射插值算法的初始值估计,计算改进算法和其它典型反马赛克方法的峰值信噪比,采用典型测试图片作为测试源进行比较.结果表明,改进算法的峰值信噪比高于其它典型插值算法的峰值信噪比,插值恢复图像的伪像也少于其它算法.改进的交替映射算法具有最好的峰值信噪比和最少的插值伪像,说明改变初始值方法可以提高交替映射算法的性能.To improve PSNR performance of alternating projection demosaicking algorithm, based on pre-estimating the minimum square error, the estimated initialization of alternating projection algorithm was meliorated. On the aspects of computed PSNR and tested typical testing pictures, the improved algorithm was compared with other typical demosaicking approaches. PSNR of the improved algorithm was better than other typical methods, and the interpolated artifact was found to be less. The improved algorithm is with best PSNR and the least interpolated artifact. The performance of alternating projection algorithm can be improved by mending the initialization.

关 键 词：颜色滤波阵列 反马赛克 预估计最小平方误差 

分 类 号：TN911.73[电子电信—通信与信息系统]