Authors:
Wael Elloumi
;
Sylvie Treuillet
;
Remy Leconge
and
Aïcha Fonte
Affiliation:
Polytech’Orléans, France
Keyword(s):
Interest Points, Local Descriptors, Matching, Performance Evolution, Video Sequence, Abrupt Motions.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
Matching Correspondence and Flow
;
Motion, Tracking and Stereo Vision
;
Real-Time Vision
;
Stereo Vision and Structure from Motion
Abstract:
In this paper, we compare the performance of matching algorithms in terms of efficiency, robustness, and computation time. Our evaluation uses as criterion, for efficiency and robustness, number of inliers and is carried out for different video sequences with abrupt motions (translation, rotation, combined). We compare SIFT, SURF, cross-correlation with Harris detector, and cross-correlation with SURF detector. Our experiments show that abrupt movements perturb a lot the matching process. They show also that SURF is the most disturbed, by such motions, and which even fails in cases that present a large rotation unlike the rest of descriptors as SIFT and cross-correlation.