A NEW SAMPLING METHOD FOR RANDOMIZED CURVE DETECTION

Abstract

In randomized curve detection, sampling is the process of collection a set of edge pexels in a edge map from which hypotheses are generated and verified. In the present work, we propose to define a sample to be a set of edge pixels on the path of a random straight line thrown into the edge map. It is shown that with the proposed sampling method, the number of random samples to take can be determined easily and reliably. This si a significant improvement over some existing works where determining this number depends one or more edge map characteristics that may not be easily obtained. Experimental results show that the proposed sampling method is robust with respect to noise and incompleteness of the curve under detction.