Title :
Complementarity Analysis of Affine Region Detectors
Author :
Yadong Sun ; Jie Yang ; Meijun Wu ; Chuan Wang
Author_Institution :
Key Lab. of Fiber Opt. Sensing Technol. & Inf. Process., Wuhan Univ. of Technol., Wuhan, China
Abstract :
Complementarity benefiting from combination of affine region detectors for image matching is examined closely. Possible combinations of five types of representative affine region detectors are integrated into the test. The complementary features and region sizes are analyzed by the means of specific instances for the first time. To objectively compare the performance of combination detectors, the matching score is used as an evaluation criterion, which reflects not only the number of correct matches, but also the ratio between correct matches and mismatches. GLOH descriptor is used for all combinations. It is observed that HA+HE, E+HE, HE+M and M+E outperform other combinational detectors in most cases.
Keywords :
image matching; GLOH descriptor; complementarity analysis; complementarity benefiting; evaluation criterion; image matching; representative affine region detector; Affine covariant regions; combination detectors; matching;
Conference_Titel :
Information Science and Engineering (ISISE), 2012 International Symposium on
Conference_Location :
Shanghai
Print_ISBN :
978-1-4673-5680-0
DOI :
10.1109/ISISE.2012.83