Searching a pattern in token scene image via multi-variant symmetric pattern matching technique

One of the fundamental tasks of pattern recognition is pattern matching. It is the act of checking for presence of a pattern's constituents within token image to have exact match. For that, most distinctive fiducial features of pattern have to be assessed and searched in the sliding windows of same pattern size formed by logically dividing the token scene image. As huge numbers of sliding windows are to be checked with pattern, pattern matching process should be time efficient and to increase pattern matching accuracy impacts due to illumination, resolution, occlusion and pose variation must be reduced. For pattern matching, this paper presents a novel local feature descriptor, multi variant symmetric local graph structure (MVSLGS) taking into account symmetric local graph structure (SLGS) as precedent approach. The computational adequacy of the proposed approach is tested on two publicly available databases with high matching accuracy, showing its proficiency over the process of pattern matching.