# Parametric Total Variation Models

Image Segmentation is often formulated as a minimization problem, where every possible labeling is assigned an energy. The idea is that the segmentation with the lowest energy corresponds to a good segmentation.

Recent research has shown that some of the minimization problems arising in image segmentation can be solved exactly. However, most energies used for segmentation depend on some external parameters in addition to the labeling. This work shows that there are some problems which can still be solved exactly. For these problems, both the optimal parameters and labeling are found.

## Source Code

Available here is some Matlab source code used for the experiments in the ICCV paper.

## Publications

- Petter Strandmark, Fredrik Kahl and Niels Chr. Overgaard,
**Optimizing Parametric Total Variation Models**International Conference on Computer Vision (ICCV), 2009. poster - Petter Strandmark, Fredrik Kahl and Niels Chr. Overgaard,
**Optimal Levels for the Two-phase, Piecewise Constant Mumford-Shah Functional**Swedish Symposium on Image Analysis - SSBA 2009*(non-refereed)*