Related Experiment Videos
A region dissimilarity relation that combines feature-space and spatial information for color image segmentation
Sokratis Makrogiannis1, George Economou, Spiros Fotopoulos
1Computer Science and Engineering Department, Wright State University, Dayton, OH 45435-0001, USA. smakrogi@cs.wright.edu
Summary
This study introduces a novel image segmentation method using cluster analysis and graph representation. It effectively integrates global feature information for improved segmentation accuracy and efficiency.
Area of Science:
- Computer Vision
- Image Processing
- Machine Learning
Background:
- Traditional image segmentation methods often struggle with integrating global feature information.
- Existing approaches may rely solely on feature-based or spatial information, limiting segmentation performance.
Purpose of the Study:
- To propose an efficient image segmentation methodology.
- To enhance segmentation by incorporating cluster analysis and graph representation principles.
Main Methods:
- Developed a feature-based, inter-region dissimilarity relation for graph-based segmentation.
- Calculated dissimilarity using proximity of region feature vectors to feature space clusters.
- Integrated global feature space information into a spatial graph representation derived from Watershed partitioning.
- Employed a region grouping process for final segmentation.
Main Results:
- The proposed method demonstrates effective image segmentation.
- Integration of global feature space information enhances segmentation results.
- Outperforms methods using only feature-based or spatial information.
Conclusions:
- The novel methodology offers an efficient approach to image segmentation.
- Combining cluster analysis and graph representation with global feature integration is effective.
- This approach provides a significant improvement over existing segmentation techniques.