Improving Image Segmentation with Agent-Based Models and Clustering
Author Information
Author(s): Erik Cuevas, Sonia Jazmín García-De-Lira, Cesar Rodolfo Ascencio-Piña, Marco Pérez-Cisneros, Sabrina Vega
Primary Institution: Universidad de Guadalajara
Hypothesis
Can integrating agent-based models with clustering methods enhance image segmentation accuracy and robustness?
Conclusion
The proposed hybrid method significantly improves image segmentation quality and robustness compared to traditional methods.
Supporting Evidence
- The hybrid method outperformed traditional segmentation methods across various quality indices.
- Agent-based preprocessing improved the uniformity of pixel intensities, enhancing clustering accuracy.
- The Firefly algorithm effectively navigated complex solution spaces for optimal clustering configurations.
- Experimental results demonstrated superior segmentation quality and robustness in noisy conditions.
Takeaway
This study shows that combining two techniques can help computers better understand and separate different parts of an image, even when the image is noisy.
Methodology
The study used an agent-based model for preprocessing to homogenize pixel intensities, followed by the Firefly metaheuristic clustering method for segmentation.
Limitations
The method may merge adjacent regions that should remain distinct, leading to a loss of critical image details.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website