Masters Theses

Author

Vikas Nahar

Abstract

"Content Based Image Retrieval System (CBIR) is used to retrieve images similar to the query image. These systems have a wide range of applications in various fields. Medical subject headings, key words, and bibliographic references can be augmented with the images present within the articles to help clinicians to potentially improve the relevance of articles found in the querying process. In this research, image feature analysis and classification techniques are explored to differentiate images found in biomedical articles which have been categorized based on modality and utility. Examples of features examined in this research include: features based on different histograms of the image, texture features, fractal dimensions etc. Classification algorithms used for categorization were 1) Mean shift clustering 2) Radial basis clustering. Different combinations of features were selected for classification purposes and it was observed that features incorporating soft decision based HSV histogram features give the best results. A library of features was then developed which can be used in RapidMiner. Experimental results for various combinations of features have also been included"--Abstract, page iii.

Advisor(s)

Erçal, Fikret
Stanley, R. Joe

Committee Member(s)

Wilkerson, Ralph W.

Department(s)

Computer Science

Degree Name

M.S. in Computer Science

Sponsor(s)

National Library of Medicine (U.S.)

Publisher

Missouri University of Science and Technology

Publication Date

Fall 2010

Pagination

xi, 84 pages

Note about bibliography

Includes bibliographical references.

Rights

© 2010 Vikas Nahar, All rights reserved.

Document Type

Thesis - Open Access

File Type

text

Language

English

Library of Congress Subject Headings

Content-based image retrieval
Image analysis -- Technique
Optical character recognition devices -- Computer programs

Thesis Number

T 9595

Print OCLC #

612405047

Electronic OCLC #

501841443

Share

 
COinS