TITLE: Color Quantization, Feature Extraction, and Content Indexing of Pathology Microscopic Images

WHEN: Wednesday, November 5, 2003 ; Time: 3:30 -5:30 PM

WHERE: Large Commons Room, 5th Fl. IS Building

WHO: Lei Zheng

Committee:

Michael J. Becich, MD PhD, Department of Pathology
Paul Munro, PhD, Information Science and Telecommunications
Hassan Karimi, PhD, Information Science and Telecommunications
John Gilbertson, MD,Department of Pathology
David Foran, PhD, Department of Pathology

ABSTRACT: After an extensive review of the state-of-the-art theories, methodologies, and implementation in information retrieval and image content indexing, a novel CBIR methodology with highly efficient adaptive color quantization based on k-means clustering, followed by discrete domain image feature extraction is proposed as basic algorithm for content-based pathology microscopic image retrieval, including the mathematical formulation of the model, detailed algorithm implementation, statistical hypothesis testing of the quality of color quantization, and performance evaluation of content-based image retrieval system. The proposed methodology is the first in its kind, and points to a new access point in solving a class of image feature extraction and content-based image retrieval problems with powerful and efficient tools in the discrete domain.