Research and Perspective on Shape Matching
形状匹配方法研究与展望

Shape matching and classification are important issues in computer vision. In recent years, contour-based shape matching approaches (e.g., shape context) and skeleton-based shape matching methods (e.g., shock graph) both have a lot of developments. In this paper, we introduce the basic concept of shape matching, give the difficulties of this topic, and provide a detailed review on the most recent approaches about shape representation and matching for both contour-based and skeleton-based methods. We also give a brief introduction about metric learning based shape retrieval. Moreover, we introduce some widely used benchmarks for shape matching in details, along with some hot topics including partial shape matching, shape classification, etc. Finally, this paper concludes with the whole framework of shape matching and the application perspective of this topic.