Neuro fuzzy and genetic algorithm for image denoising

Other Title(s)

استخدام الشبكات العصبية المضببة و الخوارزمية الجينية لإزالة الضوضاء من الصور

Dissertant

Salih, Shayma Rashid

Thesis advisor

Khudayr, Raidah Salim

Comitee Members

Rahmah, Abd al-Munim Salih
Marhun, Ali Fadil
Shaban, Hind Rustum Muhammad

University

University of Basrah

Faculty

Science College

Department

Department of Computer Science

University Country

Iraq

Degree

Master

Degree Date

2013

English Abstract

Images play an important role in many applications such as (areas of research and technology such as geographical information systems and astronomy, medical, communication applications, and etc).

Noise is introduced into images during many operations such as acquisition, transmission, compression and etc.

noise does not only cause loss of image quality but it also distorts the information storing in the image and converted it into another values (or false values).

So image denoising is important and it is the first process that must be done.

This thesis try to denoise images from several types of noise such as Gaussian, Poisson, Salt and Pepper and two types of mixed noise (Gaussian with Salt and Pepper, Poisson with Salt and Pepper).

these noises are denoised by using hybrid filters.

These hybrid filters come from combining neuro fuzzy system (which also hybrid system combining fuzzy system with neural network) with some known image filters which is mean and median filters.

In this hybrid filters, the hybridization between neuro fuzzy system and mean and median filters are depending on the type of noise that want to denoise so for denoising Gaussian or Poisson noise neuro fuzzy system will be combined with mean filter, and for denoising salt and pepper noise median filter will be used to combine with neuro fuzzy system but the value of median filter depend on the number of noisy pixels (pixels that contaminated by Salt and Pepper noise) where we use a mechanism to detect these noisy pixels and for mixed noise we use mean filter with neuro fuzzy system and also this value will be depended on the number of pixels that contaminated by Salt and Pepper noise.

When neuro fuzzy system are used as denoise filters some problem are appeared such as cannot detect the proper numbers of fuzzy set for each input and also the best input and output weights of neuro fuzzy system so genetic algorithms are used to solve these problems.

This work is programmed using MATLAB 7.7 and for evaluating the performance of hybrid filters, PSNR (peak signal to noise ratio) are used and all the results of hybrid filters will be comparing with the known image filters that used in hybridization which are mean and median filters.

The experimental results show that hybrid filters give the good results for all type of noise but genetic algorithm give the best result in Psnr and architecture.

Main Subjects

Information Technology and Computer Science

No. of Pages

111

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-744442