计算机视觉中的图匹配方法研究综述

A survey on graph matching algorithms in computer vision

图匹配是计算机视觉与模式识别领域的基础而又重要的问题.它在诸多方面都有着广泛的应用.从优化角度看,图的匹配问题是一种离散组合优化问题,使得该问题本身具有NP(non-deterministic polynomial)-hard性质.因此,寻找该问题的一种有效的近似解是当前研究的重要问题.论文首先对图匹配问题的的问题表示进行了阐述,并分析了该问题求解的难点和关键点.然后,对近年来计算机视觉研究领域中提出的一些具有代表性的传统图匹配算法进行了归纳和综述.最后,探讨了图匹配的未来研究方向和研究思路.

Graph matching is a fundamental and important issue in computer vision and pattern recognition area. It has been widely used in many computer tasks. From optimization aspect, graph matching is a discrete combinational problem and thus NP（non-deterministic polynomial）-hard. This paper first gave a brief review on graph matching problem formulation and analyzed some challenges on this problem. Then, we provided a survey on graph matching algorithms. At last, we provided suggestions on further research direction in this area.