A robust fuzzy local information C-Means clustering algorithm

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Traditional fuzzy c-means (FCM) algorithms struggle with noise and preserving image details.
• Existing FCM variations often require empirical parameter tuning, limiting their general applicability.

Purpose of the Study:

• To introduce a novel fuzzy c-means algorithm, Fuzzy Local Information C-Means (FLICM), for improved image clustering.
• To enhance clustering performance by incorporating local spatial and gray level information in a fuzzy manner.
• To develop a parameter-free algorithm that is robust to noise and preserves image details.

Main Methods:

• The proposed FLICM algorithm utilizes a novel fuzzy local similarity measure.
• It integrates both spatial and gray level information to guide the clustering process.
• The algorithm is designed to be free from empirically adjusted parameters.

Main Results:

• FLICM demonstrates superior clustering performance compared to existing FCM algorithms.
• Experiments show significant noise insensitivity and effective image detail preservation.
• The algorithm proved effective and efficient on both synthetic and real-world image datasets.

Conclusions:

• FLICM offers an effective and robust solution for image clustering, outperforming traditional FCM methods.
• The parameter-free nature and noise insensitivity make FLICM a valuable tool for image analysis.
• FLICM successfully addresses limitations of prior fuzzy clustering techniques, enhancing image processing capabilities.