Topology-Preserving Image Segmentation with Spatial-Aware Persistent Feature Matching

Bo Wen, Haochen Zhang, Dirk-Uwe G. Bartsch, William R. Freeman, Truong Q. Nguyen, Cheolhong An · 2024

Read Topology-Preserving Image Segmentation with Spatial-Aware Persistent Feature Matching by Bo Wen, Haochen Zhang, Dirk-Uwe G. Bartsch, William R. Freeman, Truong Q. Nguyen, Cheolhong An (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-02076.txt.

Description

Bo Wen, Haochen Zhang, Dirk-Uwe Bartsch, William Freeman, Truong Nguyen, and Cheolhong An wrote this paper in two thousand twenty-four. It keeps the topology of tube-like forms true when software cuts them out of biomedical images.

A cut that breaks a vessel or drops a loop can ruin later work, even if most pixels are right. Past loss rules match lasting features in a pure topology space, which can pair the wrong features. The new loss also uses where features sit in the image, so matches are guided by place. Tests on many kinds of tube images beat top rivals on topology accuracy. Code is public.

Catalog record

CREATORBo Wen, Haochen Zhang, Dirk-Uwe G. Bartsch, William R. Freeman, Truong Q. Nguyen, Cheolhong An
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-02076

Browse the full library · About the Tower of Babel