Abstract
In this paper, a new method to compute the orientability of different shapes is defined. The proposed technique is a boundary-geometry-based method that tends to take advantage of the simplicity of finding the orientability of an ellipse to obtain the orientability of any arbitrary shape. This is accomplished by finding the best-fitting ellipse of the shape. Initially, Canny edge detector is applied to obtain the edge map of the image. Convex hull points are identified and used to represent the shape. Three different approaches are presented to find the best-fitting ellipse. The three approaches use different definitions to the notion of the best-fitting ellipse of the shape. The first approach tries to find the minimum area ellipse that completely encloses the shape. While the second approach hardens the search constraints by searching for the minimum area ellipse whose center coincides with the center of the shape and completely encloses it. Alternatively, the third approach aims to find the maximum area ellipse that could be completely enclosed inside the shape and has the same center as of the shape. The three approaches utilize the particle swarm optimization technique with penalty function to solve the constrained optimization problem by defining the cost function as a multi-level function.