Content-based Image Retrieval using Multiple Shape Descriptors

(2007) Content-based Image Retrieval using Multiple Shape Descriptors. Computer Systems and Applications, 2007. AICCSA '07. IEEE/ACS International conference, 1.

[img]
Preview
PDF
14002_1.pdf

Download (18kB) | Preview
[img] Microsoft Word
14002_2.doc

Download (26kB)

Abstract

In this paper we investigate content-based image retrieval using various shape descriptors. The descriptors include 11 moment invariants, area ratios (3-concentric ring based and 8-sector based) and simple shape descriptors (eccentricity, compactness, convexity, rectangularity, and solidity). The similarity measures used are Euclidean distance and Cosine correlation coefficient. For testing, 220 binary images from SQUID categorized into 12 image groups are used. Simple Shape Descriptors with Euclidean distance achieve the best average precision (0.593). Combining simple shape descriptors and area ratios, also using Euclidean distance as similarity measure, results in 3.29% improvement.

Item Type: Article
Subjects: Computer
Department: College of Computing and Mathematics > Information and Computer Science
Depositing User: Mr. Admin Admin
Date Deposited: 24 Jun 2008 13:18
Last Modified: 01 Nov 2019 14:03
URI: https://eprints.kfupm.edu.sa/id/eprint/14002