Designing an efficient distributed digital library database for image data
Abstract
The information explosion has led to the availability of vast amounts of image data in distributed repositories. One of the goals of digital libraries is to provide efficient and effective access to this data. In this thesis, we describe how this goal can be achieved for image data. We view image databases in the context of digital libraries and provide solutions for problems which are important in that environment. Specifically, we present solutions for searching and retrieving images which can be used to support multiple groups of users with different requirements, can be used in a distributed digital library, and are scalable with respect to the size of the repository. We have addressed three phases in querying for images in a distributed digital library environment: searching the image repositories, selecting the site from which to retrieve the data, and transmitting the image to the user. To search an image database efficiently, we present uniform quantization, a feature-specific transformation for color attributes to reduce the dimensionality of the feature space which results in simpler index structures. We extend this idea to optimize queries by mapping an image query to a series of sub-queries which correspond to quantized feature vectors. We then introduce the notion of multi-level replication to support multiple levels of similarity possible between images and describe how this can be used for selecting a site for retrieval. Finally we illustrate how quantification of quality loss of images during lossy compression helps reduce the amount of data that has to be transmitted to the user.
Degree
Ph.D.
Advisors
Bhargava, Purdue University.
Subject Area
Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.