摘要
当分辨率很低时,用于图像配准的常用点特征(如角点、SIFT特征等)不明显,对应的图像配准难以正常进行。针对这一问题,本文探讨一种多尺度高能量偏移点特征配准算法,该算法以图像中偏移局部能量均值较大的点作为特征,并采用经典SIFT特征的描述方式,完成低分辨图像配准。实验结果表明,该特征稳定性好,能够有效应用于低分辨率图超分辨重建领域。
In image registration, it is very difficult to extract general point features such as corner points and SIFT features when the resolution of images is very low, and thus the associated algorithms fail to get a registra- tion result. To tackle the problem, we propose in this manuscript a novel type of point features based on multi- scale high-energy offset, a local maximum offset deviated from local energy mean, and it is called as high-energy offset （HEO） point feature. In addition, a feature descriptor and the matching algorithm adopted in SIFT fea- ture-based registration are borrowed to form HEO feature vectors and fulfill registration tasks. Experimental re- suits show that the proposed type of point features are stable and effective for registration of low resolution ima- ges.
出处
《贵州大学学报（自然科学版）》
2012年第6期91-94,共4页
Journal of Guizhou University:Natural Sciences
基金
科技部国际合作项目(No.2009DFR10530)
国家自然科学基金(No.60862003)
教育部高等学校博士点基金(No.20095201110002)
贵州省工业科技攻关项目(No.黔科合GY字(2010)3054号)
关键词
低分辨图像
图像配准
SIFT特征
角点特征
多尺度高能量偏移点特征
low resolution images
image registration
SIFT features
comer point features
multi-scale high-en-ergy offset features
