Vacuity Measure for Handwritten Character Analysis
Van Cuong Kieu, Dominique Stutzmann, Nicole Vincent · 2017
in this paper, we propose a study on the complexity measure of an object. It is based on the analysis of different details that may be limited by object contours. They may be holes or convexity evolution along the contour line. We focus in the same way on empty zones and filled zones. This study leads to a novel measure of the topology complexity - vacuity measure - that quantifies the relation between emptiness or space and objects. Based on the vacuity measure, we propose to define a novel shape descriptor and the associated dissimilarity measure. They can be applied in handwritten character analysis and in object recognition in general. The experiments are performed on a handwritten character dataset (ORIFLAMMS) and the object shape dataset (MPEG-7).