This paper addresses the issue of optimal scale selection for circular edge extraction in the context of higher dimensional multiscale edge extraction. Based on a classification of higher dimensional edges according to local curvature, we exemplarily establish a 2-D circular edge model. Through a careful mathematical derivation, we transform the circular edge model from Cartesian coordinates for which the analytical solution is unknown into polar coordinates. Utilizing this edge model we develop a novel theoretical framework for optimal scale selection for circular edge extraction through which the effects of curvature as related to scale can be analyzed. Moreover, we carry out a validation study in order to investigate on the level of principal performance how well the experimental results obtained from application of the developed framework to 2-D synthetic images match the theoretical results.
Ji-Young Lim, H. Siegfried Stiehl