Wavelet Frame-Based Fuzzy C-Means Clustering for Segmenting Images on Graphs.

Area of Science:

• Computer Vision
• Image Processing
• Graph Theory

Background:

• Traditional image processing focuses on Euclidean domains.
• Irregular image data, common in computer vision, is modeled using large graphs.
• Existing methods struggle with noise and detail preservation in graph-based image segmentation.

Purpose of the Study:

• To develop a robust fuzzy C-means (FCM) algorithm for segmenting images defined on irregular domains (graphs).
• To enhance image segmentation by incorporating spatial information and wavelet frame-based feature extraction.
• To improve noise removal and detail retention in graph-based image segmentation.

Main Methods:

• A wavelet frame-based fuzzy C-means (FCM) algorithm is proposed.
• Images on graphs are filtered using spatial information to improve robustness.
• Feature spaces are extracted using sparse approximation under tight wavelet frames.
• The FCM algorithm combines original and filtered features for segmentation of noisy images.

Main Results:

• The proposed algorithm demonstrates effectiveness and efficiency in segmenting images on graphs.
• It outperforms existing FCM-related algorithms in segmentation accuracy.
• The method successfully removes noise while retaining important feature details.
• Experimental results on synthetic and real graph images validate the approach.

Conclusions:

• The wavelet frame-based FCM algorithm offers a new, effective method for segmenting images in irregular domains.
• The approach provides superior noise reduction and detail preservation compared to existing techniques.
• This work opens new possibilities for image analysis in complex, non-Euclidean data structures.