Fast template matching based on deformable best-buddies similarity measure
Haiying Xia,Wenxian Zhao,Frank Jiang,Haisheng Li,Jing Xin,Zheng Zhou
DOI: https://doi.org/10.1007/s11042-018-6722-x
IF: 2.577
2018-10-06
Multimedia Tools and Applications
Abstract:Accuracy and speed are the essential metrics for the template matching algorithms in solving object tracking problems. Since the method based on Best Buddies Similarity (BBS) has achieved the state-of-the-art performance in terms of accuracy, matching speed becomes the shortest piece of wood of the bucket. In this paper, we propose a fast template matching method based on our deformable BBS measure. The deformable BBS measure enables matching to be performed between the patches in varying sizes, and hence leads to even higher accuracy than the original BBS-based methods. More important, we develop a fast potential-area discovery algorithm based on proposal generation and selection. It significantly reduces the numbers of useless attempts on calculating and comparing similarities of impossible image patches. The experimental results show that, with the deformable BBS measure and the fast potential-area discovery, our template matching method outperforms the state-of-the-art methods in terms of accuracy, speed and robustness.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering