Stella Yu : Publications / Google Scholar

A Unifying View of Contour Length Bias Correction
Christina Pavlopoulou and Stella X. Yu
International Symposium on Visual Computing, Las Vegas, Nevada, 30 Nov-2 Dec 2009
Paper | Slides

Original active contour formulations may becomeill-posed especially for boundaries characterized by prominent features. Attempts to yield well-posed formulations lead to bias towards short contours. We provide a framework to unify existing bias correcting energy methods and propose a novel local bias correcting scheme similar to non-maximum suppression. Our method can be seen as an approximation of a well-known algorithm that transforms a graph with positive and negative weights to a graph with only positive weights while preserving the shortest paths among the nodes.

active contour, intelligent scissors, length bias, shortest path