Related Experiment Video
Updated: Mar 31, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.7K
Segmentation of Overlapping Elliptical Objects in Silhouette Images
Summary
This study introduces a novel method for segmenting clustered, partially overlapping objects using elliptical approximations. The technique accurately identifies object contours from silhouette images, outperforming existing methods in segmentation tasks.
Area of Science:
- Computer Vision
- Image Processing
- Machine Learning
Background:
- Accurate segmentation of partially overlapping objects is crucial for numerous machine vision applications.
- Existing methods often struggle with clustered objects and require complex feature extraction.
Purpose of the Study:
- To develop an efficient method for segmenting clustered, partially overlapping objects with known elliptical shapes.
- To leverage silhouette images for robust object segmentation, requiring only foreground-background distinction.
Main Methods:
- Seedpoint extraction using bounded erosion and fast radial symmetry transform.
- Association of edge points to objects to generate contour evidence.
- Ellipse fitting to contour evidence for final contour estimation.
Main Results:
- The proposed method successfully segments partially overlapping objects using elliptical approximations.
- Experimental results on synthetic and real datasets demonstrate superior performance compared to state-of-the-art approaches.
Conclusions:
- The developed method provides an effective solution for segmenting clustered, partially overlapping objects.
- Its reliance on silhouette images and elliptical fitting offers a robust and efficient approach for machine vision.
