DECISION BASED DIRECTIONAL EDGE DETECTOR

In this work, we propose a new edge detection scheme which is called the decision based directional edge detector (DBDED). Also a modification of Cheng's shrinking algorithm is developed for producing one point edge segments. The methodology of the proposed edge detection algorithm is described in the following manner. In each of eight discrete directions, every point is analyzed in order to decide whether it is a one-dimensional (l-D) edge point in the given direction. This analysis is performed adaptively by using the calculated local directional standard deviation, local directional averages and a constant threshold. In order to prevent multiple edges, the pixels which are locally dominant in intensity are considered to be edge candidates. The true edge pixels are decided upon by eliminating some of the false edge candidates using a decision-based algorithm. It has been shown by extensive simulation work that the DBDED has satisfactory results in some preselected requirements compared with other well-known edge detection methods in the literature.