Deformable template matching using proposal-based best-buddies similarity

Publication Type:
Conference Proceeding
Citation:
Proceedings - 16th IEEE International Conference on Trust, Security and Privacy in Computing and Communications, 11th IEEE International Conference on Big Data Science and Engineering and 14th IEEE International Conference on Embedded Software and Systems, Trustcom/BigDataSE/ICESS 2017, 2017, pp. 517 - 521
Issue Date:
2017-09-07
Metrics:
Full metadata record
Files in This Item:
Filename Description Size
HaiyingXiaPaper.pdfPublished version393.89 kB
Adobe PDF
© 2017 IEEE. We propose a new method for template matching based on the Best-Buddies Similarity (BBS) measure. Our method is able to match objects with large difference in size and hence achieves a deformable template matching. In addition, compared with the original method for template matching based on the BBS, our method significantly cuts down on the computation time. The fast and deformable template matching is implemented by measuring the BBS of only potential areas instead of all positions in an image. The potential areas, which can have different size from the given template, are found by a proposal generation based on edge priors and a selective search among the obtained proposals. The results from the experiments conduct-ed on a challenging dataset demonstrate that our method out-performs the state-of-the-art methods in terms of accuracy.
Please use this identifier to cite or link to this item: