基于哈希特征的大规模图像快速匹配算法

摘要

This paper addresses the heavy image matching cost when building 3D models from large scale images. Since most images in a collection do not overlap, matching such pairs is cumbersome and does not contribute to the reconstruc-tion. In this paper a Hash-based technique is proposed to filter non-overlapping pairs to avoid exhaustive pairwise match-ing, thus improve matching efficiency. The proposed method consists of four consecutive steps:image Hash description, matching graph initialization, candidate matching graph selection and Hash matching. Experimental results show that the proposed method can greatly speed up the matching stage in 3D reconstruction.%针对基于图像进行三维重建技术在使用大规模图像集合进行重建时,需要对图像集合中图像进行两两匹配耗时问题,提出了基于哈希技术对图像构建全局哈希特征的方法,通过过滤掉无效的图像关系对来减少计算时间,极大地提高了大规模图像集合三维重建的匹配计算效率.提出的大规模图像快速哈希匹配算法包括构建图像哈希特征、构建初始匹配图、挑选候选匹配对、哈希匹配几个步骤.实验结果表明该方法能显著地提高三维重建中图像匹配的速度.