| Title | Optimizing Parametric Total Variation Models |
| Authors | Petter Strandmark, Fredrik Kahl, Niels Christian Overgaard |
| Full-text | Available as PDF |
| Alternative Location | http://www.maths.lth.se/vis... |
| Alternative Location | http://dx.doi.org/10.1109/I..., Restricted Access |
| Year | 2009 |
| Pages | 2240 - 2247 |
| Document type | Conference paper |
| Conference name | IEEE International Conference on Computer Vision (ICCV) |
| Conference Date | 2009-09-27/2009-10-04 |
| Conference Location | Kyoto, Japan |
| Status | Published |
| Quality controlled | Yes |
| Language | eng |
| Abstract English | One of the key factors for the success of recent energy<br> minimization methods is that they seek to compute global<br> solutions. Even for non-convex energy functionals, optimization<br> methods such as graph cuts have proven to produce<br> high-quality solutions by iterative minimization based on<br> large neighborhoods, making them less vulnerable to local<br> minima. Our approach takes this a step further by enlarging<br> the search neighborhood with one dimension.<br> In this paper we consider binary total variation problems<br> that depend on an additional set of parameters. Examples<br> include:<br> (i) the Chan-Vese model that we solve globally<br> (ii) ratio and constrained minimization which can be formulated<br> as parametric problems, and<br> (iii) variants of the Mumford-Shah functional.<br> Our approach is based on a recent theorem of Chambolle<br> which states that solving a one-parameter family of binary<br> problems amounts to solving a single convex variational<br> problem. We prove a generalization of this result and show<br> how it can be applied to parametric optimization. |
| Keywords | segmentation, total variation, image analysis, optimization, |
Questions: webmaster
Last update: 2013-04-11
Centre for Mathematical Sciences, Box 118, SE-22100, Lund. Telefon: +46 46-222 00 00 (vx)