Image segmentation by histogram thresholding using fuzzy sets

Area of Science:

• Image processing and computer vision.
• Digital image analysis and segmentation.
• Pattern recognition and machine learning.

Background:

• Conventional histogram thresholding methods often fail with multimodal histograms.
• Minimization of criterion functions can lead to suboptimal threshold selection.
• Local minima in thresholding algorithms limit their effectiveness for complex image data.

Purpose of the Study:

• To develop an improved histogram thresholding method for images with multimodal histograms.
• To overcome the limitations of conventional thresholding techniques, particularly local minima.
• To enhance image segmentation accuracy using a novel fuzzy similarity measure.

Main Methods:

• Proposing a histogram thresholding approach based on gray-level similarity.
• Utilizing a fuzzy measure to quantify the similarity between gray levels.
• Developing a method to circumvent local minima issues inherent in traditional algorithms.

Main Results:

• The proposed fuzzy-based histogram thresholding method effectively segments images with multimodal histograms.
• Experimental results confirm the approach's superiority over conventional methods for bimodal and multimodal cases.
• The technique successfully overcomes the problem of local minima in threshold selection.

Conclusions:

• The fuzzy measure-based histogram thresholding offers a robust solution for images with complex intensity distributions.
• This approach significantly improves image segmentation performance compared to existing methods.
• The method provides a reliable tool for accurate image analysis in various applications.