A Semiautomatic Multi-Label Color Image Segmentation Coupling Dirichlet Problem and Colour Distances.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Image segmentation is crucial for various applications, including robotics and image analysis.
• Traditional methods often struggle with accuracy and efficiency.
• Color-based segmentation offers richer information compared to intensity or texture-based methods.

Purpose of the Study:

• To develop a novel semi-automatic multi-label image segmentation technique.
• To improve segmentation accuracy and computational efficiency.
• To introduce a new color distance metric for random walk-based segmentation.

Main Methods:

• A hybrid approach combining a random walk model with direct label assignment using a novel color distance.
• Utilizing pixel probabilities from the random walker model and similarity to labeled pixels.
• Incorporating an adaptive preprocessing strategy with a regression tree for weight optimization.

Main Results:

• The proposed method demonstrated superior segmentation quality and computational time compared to state-of-the-art techniques like normalized random walk and k-means.
• Experiments on White Blood Cell (WBC) and GrabCut datasets validated the method's effectiveness.
• The approach showed robustness against noise and flexibility in color space selection.

Conclusions:

• The novel random walk and color distance-based segmentation method offers significant improvements in accuracy and efficiency.
• It provides a robust and adaptable solution for various image segmentation tasks.
• The technique shows promise for applications requiring high-quality image analysis.