利用同场景多幅图像进行图像降噪  

USING MULTIPLE IMAGES OF A SAME SCENE FOR IMAGE DENOISING

作　　者：王恒[1] 陶晓鹏[1] 

机构地区：[1]复旦大学计算机科学技术学院,上海201203

出　　处：《计算机应用与软件》2011年第8期265-269,共5页Computer Applications and Software

摘　　要：提出一种利用同场景多幅图像作为输入来降低数码照片在高感光度下数码照片噪点的方法。该方法本质上基于按区块匹配加权平均值的方法。该方法在匹配区块的搜索过程中利用了计算机视觉中的追踪、变换等技术来减小搜索范围,从而大幅度降低了整个处理过程的时间复杂度,并且增加了匹配的准确度,满足了高分辨率图像在处理速度上的需求。此外,该方法中的权值设计能够有效避免传统按区块匹配方法带来的残影问题。A method to reduce digital photo noises under high light sensitivity conditions by inputting multiple images of a same scene is proposed.The proposed method is basically based on block matching weighted average method.During the matching block searching procedure the method uses such computer vision technologies as tracking and transformation to reduce the searching area,so that it greatly reduces the time complexity of the entire processing while increasing its matching accuracy,thus satisfies the speed requirement for high resolution image processing.Meanwhile,the weight design of the method can effectively avoid the Ghost Shadow problem caused by conventional block matching methods.

关 键 词：图像降噪 多幅图像 特征追踪 仿射变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]