两步最优邻域匹配的快速图像恢复算法
Two Step Best Neighborhood Matching for Fast Image Restoration
-
摘要: 最优邻域匹配(BNM)算法运用全搜索策略,具有很高的计算复杂度.跳跃环顾最优邻域匹配(JLBNM)算法采用跳跃与环顾相结合的搜索策略,提高了匹配速度.文中提出一种两步最优邻域匹配(TSBNM)算法,利用两步搜索作为搜索策略,交替运用浏览搜索和精细搜索寻找最优匹配块;同时,精细搜索借鉴了菱形搜索算法的搜索思想,使用模板进行搜索.实验证明,在获得相当恢复质量的前提下,TSBNM算法的处理时间约为BNM算法的8%~10%,约为JLBNM算法的77%.Abstract: In contrast to best neighborhood matching (BNM) that needs intensive computation due to its full search strategy,and jump and look best neighborhood matching (JLBNM) that jointly employs jump search and look search strategy to speed up the matching procedure,two-step best neighborhood matching (TSBNM) alternatively uses browse search and fine search in block matching process.The fine search used herein employs strategy of diamond search to find matching blocks.Restoration experiments on blocks lost images show that the processing time of TSBNM is only about 8%~10% of BNM and 77% of JLBNM,meanwhile comparable restoration quality can still be obtained.
下载: