In-plane rotation and scale invariant clustering using dictionaries

Area of Science:

• Computer Vision
• Machine Learning
• Image Processing

Background:

• Content-based image retrieval (CBIR) systems require robust image clustering and feature representation.
• Existing methods often struggle with variations in image orientation and scale.

Purpose of the Study:

• To develop a novel approach for simultaneous image clustering and dictionary learning.
• To achieve in-plane rotation and scale invariant clustering for improved CBIR.

Main Methods:

• The proposed method operates in the radon transform domain.
• It simultaneously learns dictionaries and clusters images, ensuring invariance to rotation and scale.

Main Results:

• The approach demonstrated effective rotation and scale invariant clustering.
• Experiments on benchmark datasets (Smithsonian leaf, Kimia shape, Brodatz texture) showed superior retrieval performance.
• The method proved more robust than standard Gabor-based and state-of-the-art shape-based techniques.

Conclusions:

• The developed method offers a significant advancement in rotation and scale invariant image clustering.
• It provides a robust and effective solution for content-based image retrieval applications.