Award Date

1-1-2001

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Committee Member

Shahram Latifi

Number of Pages

70

Abstract

Traditionally, processing a compressed image requires decompression first. Following the related manipulations, the processed image is compressed again for storage. To reduce the computational complexity and processing time, manipulating images in the transform domain, which is possible, is an efficient solution; The uniform wavelet thresholding is one of the most widely used methods for image denoising in the Discrete Wavelet Transform (DWT) domain. This method, however, has the drawback of blurring the edges and the textures of an image after denoising. A new algorithm is proposed in this thesis for image denoising in the DWT domain with no blurring effect. This algorithm uses a suite of feature extraction and image segmentation techniques to construct filter masks for denoising. The novelty of the algorithm is that it directly extracts the edges and texture details of an image from the spatial information contained in the LL subband of DWT domain rather than detecting the edges across multiple scales. An added advantage of this method is the substantial reduction in computational complexity. Experimental results indicate that the new algorithm would yield higher quality images (both qualitatively and quantitatively) than the existing methods; In this thesis, new algorithm for image interpolation in the DWT domain is also discussed. Being different from other methods for interpolation, which focus on Haar wavelet, new interpolation algorithm also investigates other wavelets, such as Daubecuies and Bior. Experimental results indicate that the new algorithm is superior to the traditional methods by comparing the time complexity and quality of the processed image.

Keywords

Denoising; Discrete; Domain; Image; Interpolation; Transform; Wavelet

Controlled Subject

Electrical engineering

File Format

pdf

File Size

1884.16 KB

Degree Grantor

University of Nevada, Las Vegas

Language

English

Permissions

If you are the rightful copyright holder of this dissertation or thesis and wish to have the full text removed from Digital Scholarship@UNLV, please submit a request to digitalscholarship@unlv.edu and include clear identification of the work, preferably with URL.

Identifier

https://doi.org/10.25669/5puq-8ksn


Share

COinS