A Maximum-Flow Model for Digital Elastica Shape Optimization


The Elastica is a curve regularization model that integrates the squared curvature in addition to the curve length. It has been shown to be useful for contour smoothing and interpolation, for example in the presence of thin elements. In this article, we propose a graph-cut based model for optimizing the discrete Elastica energy using a fast and efficient graph-cut model. Even thought the Elastica energy is neither convex nor sub-modular, we show that the final shape we achieve is often the indeed close to the globally optimal one. Our model easily adapts to image segmentation tasks. We show that compared to previous work and state-of-the-art algorithm, our proposal is simpler to implement, faster, and yields comparable or better results.

Discrete Geometry and Mathematical Morphology - First International Joint Conference, DGMM 2021, Uppsala, Sweden, May 24-27, 2021, Proceedings
Jacques-Olivier Lachaud
Jacques-Olivier Lachaud
Professor of Computer Science

My research interests include digital geometry, geometry processing, image analysis, variational models and discrete calculus.