Wavelet based texture classification of remotely sensed images

Dissertant

al-Azraqi, Nasr Abd al-Aziz

Thesis advisor

al-Tukmashi, Salih M. Ali

Comitee Members

Naqib, Bushrah Q.
Mahdi, Ala S.

University

University of Baghdad

Faculty

College of Science

Department

Astronomy and Space Department

University Country

Iraq

Degree

Master

Degree Date

2006

English Abstract

In remote sensing, classification methods are usually improved by adopting multi-spectral-bands of satellite images.

The problem is arising when only single band is available.

In this research, the stationary wavelet transform is adopted to generate multi-band images from the available single band (i.e.

the transformed bands have the same size as the original image).

The well known and common classification methods that, usually, followed by remote sensing data users are those categorized as supervised and unsupervised methods.

Mostly, these classification techniques either clustering dependent, or statisticalfeatures depending methods.

For expert image processing scientists, it is very well known that entropy features refer to the amount of information within an image.

Our present research adopted the relative entropy of the transferred single band into multi-band images to perform the classification.

The transformation used in this research is the stationary-wavelet transforms that yield multi-same-size images.

The method is performed on different single band and multibands images, using different differentiation block sizes.

On the basis of different wavelet levels and different utilized block sizes, the obtained results are compared with the common supervised classification techniques (i.e.

minimum distance and parallelepiped classifiers).

The results proved that our presented method yields, approximately, same results as those obtained by the minimum distance classifier, using multi-bands.

The presented method has been performed by designing a Visual-basic program with a number of routines conducted to perform the classification.

Larger blobs of batches representing the image regions have been gained by our introduced method as compared with those obtained by other classification techniques (using ready software package; i.e.

ENVI).

Main Subjects

Astronomy

No. of Pages

71

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-603571