含噪图像中类线像素的检测淘汰竞争和紧支撑平滑滤波恢复算法  

Restoration Algorithm to Line Similar Pixels in Noisy Image through Detection,Wash out,Competition and Tight Frame Smoothing Filtering

作　　者：郝兵[1] 沈庭芝[1] 田卉[1] 

机构地区：[1]北京理工大学信息科学技术学院,北京100081

出　　处：《光学学报》2009年第12期3367-3373,共7页Acta Optica Sinica

基　　金：国家自然科学基金(60772066)资助课题

摘　　要：提出了一种用于检测各向异性像素集合的方向算子结构和一种非正交、非完全重构(PR)的非递推解析和合成滤波方法和算法,这种算法类似于人为组织的比赛,通过检测,淘汰,竞争和紧支撑平滑滤波,采取由粗到细,由直线到折线简化搜索和优化方案,借助少数服从多数,弱势服从强势的游戏规则,完成含噪图像中类像素的恢复任务。利用丰富的实验来检查提出算法的有效性、一致性和可靠性,得到了满意的结果。最后对算法的恢复结果作了如实的客观评价,并指出了算法的极限和不足。A new directional operator structure in uniform standard format is presented. From reference to manmade Game, through detection, wash out, competition and tight frame smoothing filtering, a non orthogonal, non perfect reconstruction and non recursive analysis and synthesis filtering macro plan and micro algorithm are proposed. With simplified and optimized strategy of coarsely detecting lines then delicately searching zigzags, and under game rules of minority obey majority and weaker follow stronger, the task of restoring line similar pixels from noisy image is completed successfully. Varity of experiments are made to check the effectiveness, consistency and reliability of proposed algorithm, and some well results are obtained. At final, some faithful objective evaluate on experimental results of new method is made, and the limitation and shortcoming of it is pointed out.

关 键 词：图像处理 方向算子 检测 淘汰 竞争 紧支撑平滑滤波 

分 类 号：TN911.73[电子电信—通信与信息系统]