Double Expansion for Optimization of Multilabel Energies

Yelena Gorelick, Zhengqin Li and Olga Veksler

Abstract

We propose a new class of energies for segmentation of multiple foreground objects with a common shape prior. Our energy involves infinity constraints. For such energies standard expansion algorithm has no optimality guarantees and in practice gets stuck in bad local minima. Therefore, we develop a new move making algorithm, we call double expansion. In contrast to expansion, the new move allows each pixel to choose a label from a pair of new labels or keep the old label. This results in an algorithm with optimality guarantees and robust performance in practice.

Session

Posters

Files

PDF iconPaper (PDF)
PDF iconSupplementary (PDF)

DOI

10.5244/C.31.19
https://dx.doi.org/10.5244/C.31.19

Citation

Yelena Gorelick, Zhengqin Li and Olga Veksler. Double Expansion for Optimization of Multilabel Energies. In T.K. Kim, S. Zafeiriou, G. Brostow and K. Mikolajczyk, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 19.1-19.11. BMVA Press, September 2017.

Bibtex

            @inproceedings{BMVC2017_19,
                title={Double Expansion for Optimization of Multilabel Energies},
                author={Yelena Gorelick, Zhengqin Li and Olga Veksler},
                year={2017},
                month={September},
                pages={19.1-19.11},
                articleno={19},
                numpages={11},
                booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
                publisher={BMVA Press},
                editor={Tae-Kyun Kim, Stefanos Zafeiriou, Gabriel Brostow and Krystian Mikolajczyk},
                doi={10.5244/C.31.19},
                isbn={1-901725-60-X},
                url={https://dx.doi.org/10.5244/C.31.19}
            }