NOVEL PERSPECTIVES ON PATTERN MATCHING WITH ASSOCIATION-BASED SYMMETRIC LOCAL FEATURES

Deep S. Dev,∗ Dakshina R. Kisku,∗∗ and Phalguni Gupta∗∗∗

Keywords

Local-feature descriptor, pattern matching, symmetric patterns

Abstract

This paper presents an efficient pattern matching method which makes use of the local-region-based light-weight feature descriptor, called Symmetric Neighbour Local Pattern (SNLP). It relies on the spatial relationship between reference pixel and its neighbours located symmetrically in horizontal, vertical and diagonal directions. SNLP is proved to be invariant to scale, illumination, image distortion and partial occlusion. Efficacy of the proposed method has been evaluated on two publically available databases and the results are found to be convincing under several challenges.

Important Links:



Go Back