Color image denoising using discrete multi wavelet

Dissertant

Imad al-Din, Matheel

University

University of Technology

Faculty

-

Department

Computer Sciences Department

University Country

Iraq

Degree

Ph.D.

Degree Date

2004

English Abstract

In many of the digital image processing applications, observed image is modeled to be corrupted by different types of noise that results in a 'noisy version.

Hence, image demising is an important problem that aims to find an estimate version from the noisy image that is as close to the original image as possible.

In the last taw years, the Multiwavelet transform has diffused in digital signal processing applications.

It plays a very important role in image processing analysis due to the fine results that it gives when used in multiresolution multi-scale modeling.

This thesis introduces firstly a proposed method of computing, one and two-dimensional multi wavelet transform using table look-up method.

Different operations in the Multiwavelet domain are introduced to modify the Multiwavelet coefficients to get the reconstructed image.

It has emerged as an efficient tool in image demising due to its effectiveness and simplicity ; Much of the work has been concentrated on, finding the best uniform threshold or best basis functions.

Secondly, several methods are proposed for the detection of noise.

A new method, for detecting noise in image called Abstract Matrix (AM) is proposed.

It is based on the second order statistics of the observed imaee.

Thirdly, several algorithms are proposed for identification of type of noise.

A new method for noise identification called Difference Matrix (DM), is proposed.

It is based on the second order statistics of the observed image.

Fourthly, many techniques are proposed for demising of gray scale and color images.

A new threshold method is proposed and compared with the other thresholding methods.

For hard thresholding, PSNR gives (26,9842 dB) value when the noise variance is (20).

The PSNR was increased in the proposed soft thresholding, it gives (27.3609 dB) PSNR value when the noise variance is (20) new tool is proposed called local multi way Ellet transforjmfiltering.

And is compared with the traditional Multiwavelet thresholding.

Two deposing methods are developed; the first is based on averaging over many shifts in the observed image.

The is based on averaging demised values of overlapping windows.

The new ods are compared with the other Multiwavelet transform deposing methods.

Their lunation with respect to many important characteristics, such as window size, isoform basis, and best shift in the observed image is achieved.

For local distinct Spinlock processing method, PSNR gives (27.1581 dB) value when the noise variance is Y20) The PSNR is increased in the local sliding processing method, it gives (28.4419 dB) PSNR value when the noise variance is (20).

Local averaging and local minimum algorithms give better PSNR results ; it gives (28.7521 dB) value for averaging and (28.8401 dB) value for minimum.

Also, other proposed algorithms introduced are called translation-invariant method and a mixture method that mixes the sliding method with the translation-invariant method.

The PSNR was increased in the translation-invariant method ; it gives (29.2915 dB) PSNR value and (29.6194 dB) PSNR value, with the mixture method when the noise variance is (20), It is concluded that Multiwavelet transform gives good results over the scalar wavelet transform.

Finally, it is shown that using Multiwavelet transform in.

image deposing provides a good quality, which makes it more suitable for deposing images applications.

Main Subjects

Information Technology and Computer Science

Topics

• Digital electronics

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-305501