A Geometric Framework for Rectangular Shape Detection

Rectangular shape detection has a wide range of applications, such as license plate detection, vehicle detection, and building detection. In this paper, we propose a geometric framework for rectangular shape detection based on the channel-scale space of RGB images. The framework consists of algorithms developed to address three issues of a candidate shape (i.e., a connected component of edge points), including: 1) outliers; 2) open shape; and 3) fragmentation. Furthermore, we propose an interestness measure for rectangular shapes by integrating imbalanced points (one type of interest points). Our experimental study shows the promise of the proposed framework.