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

67

Abstract

We introduce a wavelet domain image segmentation algorithm based on Normalized Cut (NCut) framework in this thesis. By employing the NCut algorithm we solve the perceptual grouping problem of image segmentation which aims at the extraction of the global impression of an image. We capitalize on the reduced set of data to be processed and statistical features derived from the wavelet-transformed images to solve graph partitioning more efficiently than before. Five orientation histograms are computed to evaluate similarity/dissimilarity measure of local structure. We use properties of the wavelet transform filtering to capture edge information in vertical, horizontal and diagonal orientations. This approach allows for direct processing of compressed data and results in faster implementation of NCut framework than that in the spatial domain and also decent quality of segmentation of natural scene images.

Keywords

Cut; Domain; Framework; Image Segmentation; Wavelet

Controlled Subject

Electrical engineering

File Format

pdf

File Size

1648.64 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/r5wz-w51h


Share

COinS