Tracking of convex objects

In this paper, we present a technique for grouping line segments into convex sets, where the line segments are obtained by linking edges obtained from the Canny edge detector. The novelty of the approach is twofold: first we define an efficient approach for testing the global convexity criterion, and second, we develop an optimal search based on dynamic programming or grouping the line segments into convex sets. Furthermore, we use the convexity results as the initial conditions for a deformable contour for object tracking. We show results on real images, and present a specific domain where this type of grouping can be directly applied.