Master of Science in Industrial Engineering (MSIE)
Region-based image retrieval system has been an active research area. In this study we developed an improved region-based image retrieval system. The system applies image segmentation to divide an image into discrete regions, which if the segmentation is ideal, correspond to objects. The focus of this research is to improve the capture of regions so as to enhance indexing and retrieval performance and also to provide a better similarity distance computation. During image segmentation, we developed a modified k-means clustering algorithm for image retrieval where hierarchical clustering algorithm is used to generate the initial number of clusters and the cluster centers. In addition, to during similarity distance computation we introduced object weight based on object's uniqueness. Therefore, objects that are not unique such as trees and skies will have less weight. The experimental evaluation is based on the same 1000 COREL color image database with the FuzzyClub, IRM and Geometric Histogram and the performance is compared between them. As compared with existing technique and systems, such as IRM, FuzzyClub, and Geometric Histogram, our study demonstrate the following unique advantages: (i) an improvement in image segmentation accuracy using the modified k-means algorithm (ii)an improvement in retrieval accuracy as a result of a better similarity distance computation that considers the importance and uniqueness of objects in an image.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Aulia, Eka, "Hierarchical indexing for region based image retrieval" (2005). LSU Master's Theses. 1987.