基于谱图理论的特征匹配方法研究和改进

Study and Improvement on Image Features Correspondence Based on Spectral Graph Theory

图像匹配问题是计算机视觉领域的一个基本问题,广泛地应用于很多领域,如:模式识别,自动导航,医学诊断,计算机视觉,图像三维重构等领域。将所研究的问题转化为数学问题,再利用数学工具解决这一问题,成为当今研究的一种重要手段。在这里,将图像匹配中的图像转化为数学-图论中的图,利用谱图理论解决图像匹配问题,从而形成了一类比较流行而新颖的方法,针对这一方法进行了较为系统的探究并做出了改进。

Image matching is a fundamental problem in the field of computer vision. It is widely applied in many fields such as pattern recognition,self-navigation,medical diagnosis,computer vision, 3D reconstruction and so on. In the present studies,it is an important method to convert the problems being studied into math problems and to solve them with mathematical tools. In this paper,the image in image matching into the graph in the mathematics- graph theory is converted,using spectral graph theory to solve the image matching problems,thus forming a popular and novel method. This paper provides a systematic exploration and improvement concerning this method.