Unsupervised generation of polygonal approximations based on the convex hull

Abstract

Listen

Translate article

The present paper proposes a new non-optimal but unsupervised algorithm, called ICT-RDP, for generation of polygonal approximations based on the convex hull. Firstly, the new algorithm takes into account the convex hull of the 2D closed curves or contours to select a set of initial points; secondly, the significance levels of the contour points are computed using a symmetric version of the well-known Ramer, Douglas-Peucker algorithm; and, finally, a thresholding process is applied to obtain the vertices or dominant points of the polygonal approximation. Since the convex hull can select many initial points in rounded parts of the contour, an additional deletion process is required to remove quasi-collinear dominant points. Furthermore, an additional improvement process is applied to shift the dominant points in order to increase the quality of the polygonal approximation. Experiments performed on a public available dataset show that the new proposal outperforms other unsupervised algorithms for generation of polygonal approximations.