OSN: Onion-ring support neighbors for correspondence selection

doi:10.1016/j.ins.2021.01.042

	OSN: Onion-ring support neighbors for correspondence selection
	Gong, Cheng 1; Lu, Ye 1; Song, Chunying 1; Li, Tao 1,3; Wang, Kai 1,2
	2021-06-01
发表期刊	INFORMATION SCIENCES
ISSN	0020-0255
卷号	560 页码:331-346
摘要	Correspondence Selection (CS) identifies the reliable correspondences from the putative ones. It provides fundamental cues for many computer vision tasks, therefore the performance and efficiency of CS are pivotal. Previous methods are either accurate but timeconsuming process or efficient but ambiguous. It is difficult to strike a good balance between performance and efficiency. In this paper, we propose the Onion-ring Support Neighbors (OSN) method for CS. OSN can fully exhaust the coherency information of neighbors to identify all reliable correspondences but it only requires the O(n) time and space complexity. To our best knowledge, we are the first to prove that the performance of CS can be improved by considering information from more neighbors. To this end, onion ring shaped support neighbors are introduced, whereby the directional relationships can be eliminated from the central region to neighbor regions, and rotation invariance in putative correspondences can be retained. To efficiently identify the reliable correspondences, we present the Fixed Correspondence algorithm. This algorithm can reduce the search space of putative correspondences within constant time, which contributes to a O(n) time complexity of the overall OSN process. Extensive experiments over widely used datasets show that OSN achieves state-of-the-art performance in F1-measure and Pose estimation. (C) 2021 Elsevier Inc. All rights reserved.
关键词	Correspondence selection Rotation invariance Support neighbors Coherence constraints
DOI	10.1016/j.ins.2021.01.042
收录类别	SCI
语种	英语
资助项目	National Key Research and Development Program of China[2018YFB0204304] ; National Natural Science Foundation[61872200] ; National Natural Science Foundation[62002175] ; Open Project Fund of State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences[CARCH201905] ; Open Project Fund of State Key Laboratory of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences[CARCHB202016] ; Natural Science Foundation of Tianjin[19JCZDJC31600] ; Natural Science Foundation of Tianjin[19JCQNJC00600]
WOS研究方向	Computer Science
WOS类目	Computer Science, Information Systems
WOS记录号	WOS:000641000800001
出版者	ELSEVIER SCIENCE INC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/17802
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Wang, Kai
作者单位	1.Nankai Univ, Coll Comp Sci, Tianjin, Peoples R China 2.Key Lab Med Data Anal & Stat Res Tianjin, Tianjin, Peoples R China 3.Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Gong, Cheng,Lu, Ye,Song, Chunying,et al. OSN: Onion-ring support neighbors for correspondence selection[J]. INFORMATION SCIENCES,2021,560:331-346.
APA	Gong, Cheng,Lu, Ye,Song, Chunying,Li, Tao,&Wang, Kai.(2021).OSN: Onion-ring support neighbors for correspondence selection.INFORMATION SCIENCES,560,331-346.
MLA	Gong, Cheng,et al."OSN: Onion-ring support neighbors for correspondence selection".INFORMATION SCIENCES 560(2021):331-346.