Development and gray scale dynamic programming texture classification based on wavelet network

مقدم أطروحة جامعية

Hasan, Hasan Jalil

الجامعة

الجامعة التكنولوجية

الكلية

-

القسم الأكاديمي

قسم هندسة السيطرة و النظم

دولة الجامعة

العراق

الدرجة العلمية

ماجستير

تاريخ الدرجة العلمية

2004

الملخص الإنجليزي

Image identification plays a great role in industrial, remote sensing, and military applications.

It is concerned with the generation of a signature to the image.

This thesis proposes a dynamic programming algorithm for texture classification.

It was based on a new mapping network called wavelet-network (WN) WN was employed to make approximation to the images before their passing through the discrete wavelet transform decomposition to extract image descriptive features.

These features are used in the proposed image identification algorithm.

The conventional Wavelet Transform with Minimum Distance (WTMD) approach is employed here also for image identification.

A successful identification rate of 99 % was achieved with this approach.

Sub-image identification algorithm has been developed to make it possible to find out the given subpart of an image, this algorithm is also implemented using WTMD after subdividing each image into four subparts and dealing with, each subpart as a unique image.

A successful identification rate of 96.4% was achieved with this approach.

In this thesis the neural network (NN) classifier is combined with the wavelet transform.

As a result, another approach of identification is introduced, namely the Wavelet Transform with Neural Network (WNN), the architecture of the proposed neural network is a Feed-Forward Multilayer Perceptron.

It uses the back-propagation in the training phase.

The identification rate of using WTMD was improved to 100 % after employing WNN.

A reference set of 100 images is used, collected from different texture of.

التخصصات الرئيسية

تكنولوجيا المعلومات وعلم الحاسوب

الموضوعات

• الإلكترونيات الرقمية

• APA
• MLA
• AMA

لغة النص

الإنجليزية

نوع البيانات

رسائل جامعية

رقم السجل

BIM-305992