Superpixel-Based Image Segmentation Using Squared 2-Wasserstein Distances
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918303550668800 |
|---|---|
| author | Huang, Jisui Alpers, Andreas Chen, Ke Lei, Na |
| author_facet | Huang, Jisui Alpers, Andreas Chen, Ke Lei, Na |
| contents | We present an efficient method for image segmentation in the presence of strong inhomogeneities. The approach can be interpreted as a two-level clustering procedure: pixels are first grouped into superpixels via a linear least-squares assignment problem, which can be viewed as a special case of a discrete optimal transport (OT) problem, and these superpixels are subsequently greedily merged into object-level segments using the squared 2-Wasserstein distance between their empirical distributions. In contrast to conventional superpixel merging strategies based on mean-color distances, our framework employs a distributional OT distance, yielding a mathematically unified formulation across both clustering levels. Numerical experiments demonstrate that this perspective leads to improved segmentation accuracy on challenging images while retaining high computational efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_17071 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Superpixel-Based Image Segmentation Using Squared 2-Wasserstein Distances Huang, Jisui Alpers, Andreas Chen, Ke Lei, Na Computer Vision and Pattern Recognition Probability We present an efficient method for image segmentation in the presence of strong inhomogeneities. The approach can be interpreted as a two-level clustering procedure: pixels are first grouped into superpixels via a linear least-squares assignment problem, which can be viewed as a special case of a discrete optimal transport (OT) problem, and these superpixels are subsequently greedily merged into object-level segments using the squared 2-Wasserstein distance between their empirical distributions. In contrast to conventional superpixel merging strategies based on mean-color distances, our framework employs a distributional OT distance, yielding a mathematically unified formulation across both clustering levels. Numerical experiments demonstrate that this perspective leads to improved segmentation accuracy on challenging images while retaining high computational efficiency. |
| title | Superpixel-Based Image Segmentation Using Squared 2-Wasserstein Distances |
| topic | Computer Vision and Pattern Recognition Probability |
| url | https://arxiv.org/abs/2601.17071 |