A Study on Continuous Max-Flow and Min-Cut Approaches

Abstract. This work addresses a class of total-variation based multilabeling problems over a spatially continuous image domain, where the data fidelity term can be any bounded function, not necessarily convex. The feasible label values can be constrained to any discrete finite set. In the spatially discrete setting, Ishikawa [31] showed that such labeling problems can be solved exactly by standard max-flow and min-cut algorithms over specially designed graphs. We will propose a continuous analogue of Ishikawa’s graph construction [31] by formulating new continuous max-flow and min-cut models over a specially designed domain. These max-flow and min-cut models are equivalent under a primal-dual perspective. They can be seen as exact convex relaxations to the original problem and can be used to compute global solutions. The continuous max-flow and min-cut models can also be extended to problems with continuous label values. In the end, we propose fast continuous max-flow based al...

• Type: Conference
• Year: 2010
• Where: CVPR
• Authors: Jing Yuan, Egil Bae, Xuecheng Tai
• Keywords: Computer Vision, Continuous Image Domain, Continuous Max-flow, CVPR 2010, Min-cut Models