PicToSeek: combining color and shape invariant features for image retrieval.

Area of Science:

• Computer Vision
• Image Processing
• Information Retrieval

Background:

• Current image retrieval systems often struggle with variations in illumination, object pose, and clutter.
• Developing robust and accurate methods for object recognition and retrieval remains a key challenge in computer vision.

Purpose of the Study:

• To develop and evaluate a novel image retrieval system that combines color and shape invariant features.
• To enhance object retrieval accuracy and robustness by integrating complementary invariant features.

Main Methods:

• Proposed color models independent of object geometry, pose, and illumination.
• Derived color invariant edges and computed shape invariant features.
• Combined color and shape invariants into a unified high-dimensional feature set for object retrieval.

Main Results:

• Object retrieval using combined color and shape invariants achieved excellent accuracy.
• Color invariants alone provided very high retrieval accuracy.
• Shape invariants alone showed poor discriminative power.
• The retrieval scheme demonstrated robustness to partial occlusion, object clutter, and changes in object pose.

Conclusions:

• Composite color and shape invariant features offer a powerful approach for accurate and robust image retrieval.
• The developed image retrieval scheme is effective for searching multicolored man-made objects in real-world scenes.
• The system has been integrated into the PicToSeek system for web-based image searching.