Related Experiment Video
Updated: Jul 7, 2026
10:39
A Label-Free Segmentation Approach for Intravital Imaging of Mammary Tumor Microenvironment
Published on: May 24, 2022
Automatic image segmentation by integrating color-edge extraction and seeded region growing
J Fan1, D Y Yau, A K Elmagarmid
1Dept. of Comput. Sci., Purdue Univ., West Lafayette, IN 47907, USA.
Summary
This study introduces an automatic image segmentation method using color edges and seeded region growing (SRG). The technique accurately identifies object boundaries and is applied to face detection and human object generation.
Area of Science:
- Computer Vision
- Image Processing
- Artificial Intelligence
Background:
- Automatic image segmentation is crucial for various computer vision tasks.
- Existing methods often struggle with accurate boundary detection and object identification.
Purpose of the Study:
- To develop a novel automatic image segmentation method.
- To enhance the accuracy of boundary detection in segmented images.
- To apply the method for automatic face detection and human object generation.
Main Methods:
- Color edges are detected using an improved isotropic edge detector and entropic thresholding.
- Seeded region growing (SRG) is initiated using centroids of edge regions.
- Integration of color-edge extraction and SRG for precise boundary delineation.
- Seeded region aggregation is used for semantic object generation.
Main Results:
- The proposed method generates homogeneous image regions with accurate and closed boundaries.
- Successful application of the segmentation method to automatic face detection.
- Generation of semantic human objects from detected faces.
Conclusions:
- The developed image segmentation technique offers improved accuracy in boundary detection.
- The method shows potential for practical applications in face detection and object recognition.
