Related Experiment Video
Updated: Jun 20, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
Unsupervised multiphase segmentation: a phase balancing model
Berta Sandberg1, Sung Ha Kang, Tony F Chan
1Adel Research, Inc., Los Angeles, CA, USA. berta.sandberg@adelresearch.com
Summary
This study introduces a new unsupervised multiphase image segmentation model that automatically determines the optimal number of phases. The model uses a novel regularization term for robust and stable image segmentation results.
Area of Science:
- Computer Vision
- Image Processing
- Mathematical Modeling
Background:
- Variational models have been a cornerstone of image segmentation since the Mumford-Shah functional.
- Existing methods often require pre-defined phase numbers, limiting their adaptability.
Purpose of the Study:
- To develop an unsupervised multiphase image segmentation model.
- To introduce a novel regularization term that enables automatic phase number selection.
- To demonstrate the robustness and stability of the proposed segmentation approach.
Main Methods:
- Proposed a new variational functional for multiphase image segmentation.
- Incorporated a phase scale measure as a regularization term.
- Developed a fast, brute-force numerical algorithm for model implementation.
Main Results:
- The model automatically selects an optimal number of phases during segmentation.
- Experimental results demonstrate the robustness and stability of the proposed unsupervised segmentation model.
- The intensity fitting term drives segmentation, complemented by the phase scale regularization.
Conclusions:
- The proposed model offers an effective unsupervised approach to multiphase image segmentation.
- Automatic phase number selection enhances model adaptability and user-friendliness.
- The model's stability and robustness are validated through experimental evidence.
