On segmentation by total variation type energies of Kobayashi-Warren-Carter type with fidelity
Yoshikazu Giga, Ayato Kubo, Hirotoshi Kuroda, Jun Okamoto, Koya Sakakibara

TL;DR
This paper introduces a novel total variation energy derived as a limit of Kobayashi-Warren-Carter energy, analyzing its minimizers and segmentation properties, especially in one-dimensional cases with continuous data.
Contribution
It establishes the piecewise constant nature of minimizers for this energy and provides bounds on the number of jumps, differing from classical TV models.
Findings
Minimizers are piecewise constant with bounded jumps in 1D.
Number of segments is explicitly bounded by data fidelity constants.
Existence of minimizers is proven in multi-dimensional bounded data cases.
Abstract
We consider a total variation type energy which measures the jump discontinuities different from usual total variation energy. Such a type of energy is obtained as a singular limit of the Kobayashi-Warren-Carter energy with minimization with respect to the order parameter. We consider the Rudin-Osher-Fatemi type energy by replacing relaxation term by this type of total variation energy. We show that all minimizers are piecewise constant if the data function in the fidelity term is continuous in one-dimensional setting. Moreover, the number of jumps is bounded by an explicit constant involving a constant related to the fidelity. This is quite different from conventional Rudin-Osher-Fatemi energy where a minimizer has no jumps if the data has no jumps. Our results give an upper bound of the number of segments in a segmentation problem. The existence of a minimizer is guaranteed in…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
