Feature Extraction And Clustering Of Trademark Images: A Comparison Between Geometric Moment Invariant And Zernike Moment Invariant

ABD HALIM, SHAHLIZA (1999) Feature Extraction And Clustering Of Trademark Images: A Comparison Between Geometric Moment Invariant And Zernike Moment Invariant. Masters thesis, Universiti Teknologi Malaysia.

[img]
Preview
PDF
shahliza_abd_halim(master).PDF

Download (3441Kb) | Preview

Abstract

An ever increasing numbers of registered trademarks have made trademarks classification for the registration purpose, even more tedious. A tool is needed to automate the matching of trademark images by comparing a new trademark image with existing trademark images in the repository.A suitable feature extraction is essential in order to extract unique invariant features from these trademarks. In this study, a comparison is made between two feature extraction techniques,Zernike Moment Invariant and Geometric Moment Invariant. Comparison is based on the criteria of "good features" which are interclass and intraclass sensitivity. The results of this study indicate that Zernike Moment Invariant technique performs better in terms of interclass sensitivity when compared to Geometric Moment Invariant technique. On the other hand, Geometric Moment Invariant technique has a slightly better intraclass sensitivity compared to Zernike Moment Invariant features. The results are analyzed based on plotted graphs and on the map of clusters produced by the Self Organizing Maps (SOM)

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Faculty of Computer Science and Information System > Software Engineering
Depositing User: Unnamed user with email knizam@utm.my
Date Deposited: 04 Jul 2013 07:21
Last Modified: 04 Jul 2013 07:21
URI: http://ir.fsksm.utm.my/id/eprint/3045

Actions (login required)

View Item View Item