A Two-Stage Automatic Color Thresholding Technique.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Image thresholding is crucial for computer vision, enabling focus on relevant objects by suppressing background noise.
• Existing methods often require training data or struggle with varying conditions, limiting their real-world application.

Purpose of the Study:

• To develop an unsupervised, automated background suppression technique for enhanced image analysis.
• To improve the accuracy and robustness of foreground-object segmentation in diverse imaging scenarios.

Main Methods:

• A novel two-stage histogram-based approach utilizing pixel chromaticity for background suppression.
• The method operates without requiring training or ground-truth data, ensuring full automation.

Main Results:

• Demonstrated robust background-foreground separation on printed circuit assembly (PCA) boards and skin cancer lesion datasets.
• Achieved superior performance compared to existing state-of-the-art thresholding methods under varied lighting and camera conditions.

Conclusions:

• The proposed method offers a significant advancement in automated image analysis, particularly for tasks involving small objects or subtle features.
• Its unsupervised and robust nature makes it suitable for real-world applications in industrial inspection and medical diagnostics.