Shape Based Image Retrieval
Michael Brown
Department of Computer Science
University of Canterbury
Abstract
A new technique is proposed for representing shape features for the purpose of image retrieval. This project defines the properties of this representation, and implements software that extracts the relevant features from a given image and converts them into a recognised format. The project also implemented testing software that enables queries to be performed on a medium sized database of images and evaluates the performance of the proposed representation. The representation achieved 82% precision and 83% recall.
