基于Delaunay三角化的形状表示与相似性衡量(英文)

Shape Representation and Similarity Measure Based on Delaunay Triangulation

在计算机视觉中,形状的表示和相似性衡量是重要且复杂的问题,提出了一种改进的SUSAN(最小一致性区域)拐点检测算法并用于形状表示,同时基于Delaunay三角化给出了一个用于形状相似性衡量的有效算法。首先,对形状的拐点进行Delaunay三角形构造,然后从Delaunay三角网中获得Delaunay图矩阵,最后使用矩阵的谱对拐点进行匹配。在含有1 400幅图像的MPEG-7 CE-Shape-1数据库中的检索实验进一步验证了算法的有效性。

Shape representation and similarity measure are important and difficult problems in computer vision and have been extensively studied for decades. This paper presents an enhanced SUSAN （Smallest Univalue Segment Assimilating Nucleus） Corner Detector for shape representation and an effective algorithm to establish shape similarity measure based on Delaunay triangulation. Firstly, delaunay triangulation was constructed among corners of each shape which has been normalized in advance. Secondly, the Delaunay graph matrix was achieved from Delaunay triangulation net. Finally, the corners were matched by using spectrum of the graph matrix. Shape retrieval Experiments have been conducted on the MPEG-7 Core Experiment CE-Shape- 1 database of 1 400 images which illustrate good performance of the algorithm.