Fast 2D Shape Matching with Centroid Shape Context in Polar Configuration

摘要

In this paper, a simple yet fast centroid shape context is proposed for fast 2D shape matching. To improve the efficiency, centroid shape context is defined to capture the distribution of all points directly based on the centroid of object contour in polar configuration. With this descriptor, a sliding matching method based on strategy of dynamic programming is proposed to enhance the computational efficiency and invariance to rotation between two shapes during shape matching. Experiments on public shape benchmarks (e.g., MPEG-7 dataset, Kimia's dataset and ETH-80 dataset) demonstrate the efficiency of the proposed algorithm compared with the state-of-the-art shape matching approaches.