Elastica Energy Regularization via Graph Cuts

We propose a graph cut model to optimize bidimensional shapes with respect to the elastica energy. At each iteration our model selects the shape of minimum elastica value among a set of candidates generated by a discrete process that we call the balance coefficient flow. In this work we show how the balance coefficient flow relates with the curve-shortening flow and how our model can be included in an image segmentation pipeline. Finally, we provide a study to evaluate the effects of our model in the image segmentation task.

Research report, hal-04421328, 2024