A hierarchical approach to color image segmentation using homogeneity

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Color image segmentation is crucial for image analysis.
• Existing methods often struggle with local variations and small objects.
• Hierarchical approaches offer potential for improved segmentation accuracy.

Purpose of the Study:

• To develop a novel hierarchical approach for color image segmentation.
• To enhance region identification using homogeneity histograms and hue analysis.
• To improve robustness against local variations and image singularities.

Main Methods:

• Extending histogram concept to the homogeneity domain for multilevel thresholding.
• Analyzing hue feature within uniform regions identified in the first phase.
• Employing a region merging process to prevent over-segmentation.
• Utilizing CIE(L*a*b*) color space for color difference measurement.

Main Results:

• Successfully identified uniform regions using multilevel thresholding on homogeneity histograms.
• Reduced image singularities by approximately 99.7% through hue value redefinition.
• Demonstrated effective handling of small objects and local color variations.
• Achieved superior performance in hierarchical segmentation across extensive color image datasets.

Conclusions:

• The proposed hierarchical method provides an effective and superior approach to color image segmentation.
• The integration of homogeneity histograms and hue analysis significantly improves segmentation quality.
• The method demonstrates robustness in handling image complexities and reducing singularities.