Corners and edges are all important image features in many vision-based areas. Corners are more reliable than edges and much easier matched because of their sparseness, while edges contain richer scene structure-information more applicable of 3D recognition. A new fast and robust edge-matching algorithm based on matched corners is proposed. In the matching process, corner constraint and edge constraint are introduced. Firstly, the matched corners are used to guide the edge matching. How to use the previously matched corners to guide and constrain edge matching is presented. Furthermore, propagation idea is introduced to get matched edges. Secondly, edge constraint is proposed to limit the search area in several pixels, then epipolar constraint is also used to achieve the matched points, if necessary the correlation score will be utilized. Numerous experiments with various real images clearly show that if the two images differences are not too severe, the benefit of integrating corner matches into the matching procedures is obvious, and the algorithm greatly improves the speed as well as the correct matching ratio to higher than 97%.

Date Published: 28 October 2006
PDF: 6 pages
Proc. SPIE 6358, Sixth International Symposium on Instrumentation and Control Technology: Sensors, Automatic Measurement, Control, and Computer Simulation, 635815 (28 October 2006); doi: 10.1117/12.717813

Published in SPIE Proceedings Vol. 6358:
Sixth International Symposium on Instrumentation and Control Technology: Sensors, Automatic Measurement, Control, and Computer Simulation
Jiancheng Fang; Zhongyu Wang, Editor(s)