The Four Color Theorem for Cell Instance Segmentation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Ye, Zhou, Yu, Wang, Yifeng, Xiao, Jun, Wang, Ziyue, Zhang, Yongbing, Chen, Jianxu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908404507738112
author Zhang, Ye
Zhou, Yu
Wang, Yifeng
Xiao, Jun
Wang, Ziyue
Zhang, Yongbing
Chen, Jianxu
author_facet Zhang, Ye
Zhou, Yu
Wang, Yifeng
Xiao, Jun
Wang, Ziyue
Zhang, Yongbing
Chen, Jianxu
contents Cell instance segmentation is critical to analyzing biomedical images, yet accurately distinguishing tightly touching cells remains a persistent challenge. Existing instance segmentation frameworks, including detection-based, contour-based, and distance mapping-based approaches, have made significant progress, but balancing model performance with computational efficiency remains an open problem. In this paper, we propose a novel cell instance segmentation method inspired by the four-color theorem. By conceptualizing cells as countries and tissues as oceans, we introduce a four-color encoding scheme that ensures adjacent instances receive distinct labels. This reformulation transforms instance segmentation into a constrained semantic segmentation problem with only four predicted classes, substantially simplifying the instance differentiation process. To solve the training instability caused by the non-uniqueness of four-color encoding, we design an asymptotic training strategy and encoding transformation method. Extensive experiments on various modes demonstrate our approach achieves state-of-the-art performance. The code is available at https://github.com/zhangye-zoe/FCIS.
format Preprint
id arxiv_https___arxiv_org_abs_2506_09724
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Four Color Theorem for Cell Instance Segmentation
Zhang, Ye
Zhou, Yu
Wang, Yifeng
Xiao, Jun
Wang, Ziyue
Zhang, Yongbing
Chen, Jianxu
Computer Vision and Pattern Recognition
Cell instance segmentation is critical to analyzing biomedical images, yet accurately distinguishing tightly touching cells remains a persistent challenge. Existing instance segmentation frameworks, including detection-based, contour-based, and distance mapping-based approaches, have made significant progress, but balancing model performance with computational efficiency remains an open problem. In this paper, we propose a novel cell instance segmentation method inspired by the four-color theorem. By conceptualizing cells as countries and tissues as oceans, we introduce a four-color encoding scheme that ensures adjacent instances receive distinct labels. This reformulation transforms instance segmentation into a constrained semantic segmentation problem with only four predicted classes, substantially simplifying the instance differentiation process. To solve the training instability caused by the non-uniqueness of four-color encoding, we design an asymptotic training strategy and encoding transformation method. Extensive experiments on various modes demonstrate our approach achieves state-of-the-art performance. The code is available at https://github.com/zhangye-zoe/FCIS.
title The Four Color Theorem for Cell Instance Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.09724