PATTERN MATCHING USING ALL-DIRECTION SYMMETRIC STRUCTURE WITH IMAGE VARIATIONS

Deep S. Dev and Dakshina R. Kisku

Keywords

Pattern matching, sliding windows, local-region feature descriptor, all-direction symmetric LGS

Abstract

Pattern matching aims to search for the pattern components in the image to look for precise similarity. However, the pattern matching is affected due to changes in orientation, resolution, illumination and occlusion in the image. This article presents a novel local descriptor called all-direction symmetric local graph structure (AdSLGS) for pattern matching. AdSLGS is invariant to scale and illumination, which extracts features in a more symmetric way with respect to local binary pattern, local graph structure and symmetric local graph structure. It increases the accuracy of pattern matching in contrast to other processes while image variations are observed, and simultaneously, it reduces time and memory requirements by the computing system. The proposed pattern matching algorithm is tested on two benchmark databases, namely, COIL-100 and Caltech-101 with a high note of matching accuracy, exhibiting its robustness in the process in the presence of challenging image constraints.

Important Links:

Go Back