Convexity problems on reconfigurable meshes

The reconfigurable mesh, a parallel system with bus autonomy, can support various interconnection schemes during execution of an algorithm. It offers very efficient computation power in many application domains. In digital image processing and computer vision, convexity is a natural shape descriptor and a classifier for objects in the image space. We first present that the problem of identifying extreme points of convex hulls can be solved in O(1) time on the reconfigurable mesh proposed. Furthermore, we present constant time algorithms for a number of convexity-related problems on reconfigurable meshes. These problems include point inclusion, interior detection, area, and width of convex hulls.