一种基于SURF和扩展哈希的空间约束图像匹配算法  

A Spatial Constraint Image Matching Algorithm Based on SURF and Extended Hash

作　　者：吴铭心[1] 

机构地区：[1]北京航空航天大学高等教育研究所,北京100191

出　　处：《重庆师范大学学报（自然科学版）》2015年第2期104-110,共7页Journal of Chongqing Normal University:Natural Science

摘　　要：图像匹配是计算机视觉中一个重要的研究方向,是图像拼接、图像检索等相关应用的基础工作。如何实现快速、高效的图像匹配技术是本文的主要研究内容。提出一种基于SURF和扩展哈希的空间约束图像匹配算法,为了提高特征检测的速度,首先提取SURF特征描述图像局部特征,然后在局部敏感性哈希算法基础上,提出一种改进的高维数据搜索算法,该改进算法变换局部敏感性哈希的投影空间,使变换后的每一维特征数据都比原算法具有更好的局部敏感性。最后采用空间约束RANSAC算法剔除误匹配点,进一步增加算法的鲁棒性。实验结果表明,本文提出的算法与传统算法如BBF、LSH以及iDistance等算法相比具有更优的搜索效率,在一定程度上提高了图像匹配的性能。Image matching is an important research in computer vision, which also is a fundamental task for image stitching, image retrieval and other computer vision application areas. The main work of this paper is to find image matching technology with speedi- ness and high efficiency. The paper presents a spatial constraint image matching algorithm based on SURF and extended hash. The local invariant feature SURF descriptors are extracted from images, which greatly improves speed of feature detection. Then a high dimensional search algorithm is improved inspired by local sensitive hash algorithm （LSH）, which enhances the local sensitivities of local sensitive hash by changing the projection space. At last, the mismatching points are wiped out by the RANSAC algorithm in order to enhance the robustness of the algorithm. The experiment results express that the algorithm obtains higher search accuracy and efficiency comparing to those classical high dimensional search methods such as BBF, LSH, iDistance and so on, and obviously improves the performance of image matching.

关 键 词：图像匹配 SURF算法 局部敏感哈希 特征搜索 投影空间 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]