基于不变特征的区域相关技术  

Area Correlation Using the Invariant Feature

作　　者：王亚东[1] 彭嘉雄[1] 魏智[1] 

机构地区：[1]华中理工大学图像识别与人工智能研究所

出　　处：《华中理工大学学报》1996年第2期4-6,共3页Journal of Huazhong University of Science and Technology

基　　金：航天基础性研究基金

摘　　要：研究了如何利用不变特征实现区域相关；给出了具有平移、旋转、比例不变性的两组特征，即不变矩特征和同心圆特征；提出了计算二值图像不变矩的快速算法．通过对复杂背景下的区域相关实验，证实了利用不变特征相关的可行性．rea correlation algorithms are usually performed using image intensity pixels by pixels,which are not only time-consuming but also lack of robustness to geometrical distortion. A study ismade on an area correlation algorithm using invariant features. Two groups of features, i. e., invariant moments features and concentric circles features, which are invariant in translation, rotation andscale, are proposed. A fast algorithm for invariant moments in binary images is given. The feasibilityof the correlation algorithm is proved by correlation experiments with complex background images.

关 键 词：区域相关 不变特征 不变矩 二值图像 遥感图像 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]