Histogram thresholding using fuzzy and rough measures of association error.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Traditional histogram thresholding methods often require prior assumptions about image histograms.
• Existing techniques may struggle with ambiguous regions or complex image structures.

Purpose of the Study:

• To propose a novel histogram thresholding methodology utilizing fuzzy and rough set theories.
• To develop a technique that does not require prior assumptions about the histogram.
• To introduce a robust quantitative index for evaluating image segmentation performance.

Main Methods:

• Bilevel thresholding by associating histogram elements with regions based on association errors derived from fuzziness and roughness.
• Multilevel thresholding implemented using a tree-structured algorithm based on the proposed bilevel method.
• Development of a quantitative evaluation index using the median of absolute deviation from the median.

Main Results:

• Successful application of the methodology for image segmentation, object/background separation, and edge extraction.
• Demonstrated effectiveness through extensive qualitative and quantitative experimental results.
• The proposed method shows robustness and accuracy in image analysis tasks.

Conclusions:

• The proposed fuzzy and rough set-based histogram thresholding offers a powerful and assumption-free approach to image segmentation.
• The novel quantitative index provides a reliable measure for assessing segmentation performance.
• The methodology is effective for various image processing tasks, including object separation and edge detection.