An Improved Compatibility Based on Weighted Local and Global Information for Graph Matching

Abstract

Graph matching is a fundamental NP-problem in computer vision and pattern recognition. The computing of compatibility is a key problem in graph matching. In order to overcome the limitation of Euclidean distance when computing the compatibility, we introduce the weighted local and global topology structure information as a second order compatibility term. We have tested our method on a group of synthetic graphs, the CMU house sequence and the real images. Experimental results demonstrate the superior performance of our weighted compatibility based on the local and global topology structure information as compared with the traditional method based on the Euclidean distance.