Integrating fast fourier transformation with constrained independent component analysis for facial image retrieval

Other Title(s)

دمج سريع التحول فورييه و مقيدة تحليل المركبات المستقلة لاسترجاع صور الوجه

Dissertant

Abu Seilik, Umar Awdah Ahmad

Thesis advisor

Samawi, Venus W.

Comitee Members

al-Nihoud, Jihad Quball Awdah
Bani Muhammad, Sad
Slait, Azzam Talal

University

Al albayt University

Faculty

Prince Hussein Bin Abdullah Faculty for Information Technology

Department

Department of Computer Science

University Country

Jordan

Degree

Master

Degree Date

2013

English Abstract

Content-based image retrieval (CBIR) is a method of searching, browsing, and querying images based on image content.

In this work, we focus on the development of a content-based facial image retrieval technique based on the Constrained Independent Component Analysis (CICA). Recently, Independent Component Analysis (ICA) in reduced Principle Component Analysis (PCA) space is used for characterizing query image.

ICA is used to identify as many ICs as number of observations.

There is an arbitrary ordering of the extracted ICs.

To overcome arbitrary ordering of the extracted ICs, CICA was developed so that only the desired ICs are retrieved. From previous researches, it was found that using PCA with ICA improves the recognition rate.

Therefore, we found it worthwhile to study the effect of using PCA with CICA.

CICA algorithm has some limitations concerning its sensitivity to scale, rotation, and translation.

To overcome the CICA limitations, Fast Fourier Transform (FFT) will be applied on the image before applying CICA. In this work, we aim to study the effect of using edge detection, PCA, and FFT with CICA in content based image retrieval (CBIR) for facial image (as specific domain data).

We suggest four different approaches, using PCA to reduce dimensionality, edge detection to reduce noise, and FFT to overcome some CICA sensitivity problem.

The four approaches are: combing PCA with CICA, perform edge detection then apply PCA and CICA, apply FFT the PCA and CICA, and final approach combines edge detection, FFT, PCA and CICA.

A comparison between the four approaches is performed to specify the recommend approach (the one with heights recognition and retrieving ability).

The comparison between these approaches is from recognition accuracy, average recall and average precision point of views.

The results indicate the effect of FFT2, PCA, and edge detection on the system accuracy.

The experimental results of the proposed CBIR system is tested with public facial databases, the ORL and Yale face database.

The best Recognition accuracy using ORL database is 95.3 %, and the best recognition accuracy using Yale database is 96 %.

By comparing the recognition accuracy of our recommended approach with the accuracy of other works (works that used the same databases), it is evident that our approach outperforms these work.

Main Subjects

Information Technology and Computer Science

Topics

• Digital electronics
• Indexing

No. of Pages

55

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-322006