Applying information theory to image segmentation

Pao Yue-kong Library Electronic Theses Database

Applying information theory to image segmentation

 

Author: To, Siu-wah
Title: Applying information theory to image segmentation
Degree: M.Sc.
Year: 1999
Subject: Image processing -- Digital techniques
Hong Kong Polytechnic University -- Dissertations
Department: Multi-disciplinary Studies
Dept. of Electronic Engineering
Pages: [119] leaves : ill. ; 31 cm
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b1471966
URI: http://theses.lib.polyu.edu.hk/handle/200/409
Abstract: Image segmentation is the process to separate an image into its constituent parts, or objects. There are many different segmentation algorithms but there is no consensus on what criterion is to be used to evaluate the performance of a particular segmentation algorithm. In this project, we shall make use of a new evaluation criterion, which is based on the information content of a segmented image, to evaluate the quality of a segmented image. We call this "Segmented Image Entropy" (SIE). A good segmented image should contain a large amount of information, and hence a large SIE. Similarly, we also measure the information content of the input gray scale image with a criterion called "Gray Scale Image Entropy" (GIE). With the use of SIE and GIE, the performance of different segmentation algorithms will be evaluated. Several common segmentation algorithms, namely Otsu's method, Maximum Entropy method, Minimum Error Thresholding, Minimum Cross Entropy method, Maximum Segmented Image Information Thresholding were evaluated and compared. According to a bi-Guassian p.d.f. model, 1400 gray scale images were generated with different combination in parameters, such as object size, mean and standard deviation of the object pixel gray-level values, mean and standard deviation of the background pixel gray-level values. Their segmented images were obtained by applying these chosen segmentation algorithms. Then, we calculated the GIE and SIE for each image. These 1400 (GIE, SIE) pairs are used in the comparison of the performance of different segmentation algorithms. Among these 1400 images, some are with good input quality but the segmentation results are not good. So, we can identity the weaknesses of a segmentation algorithm and improve it. For a particular segmentation algorithm called Maximum Segmented Image Information Thresholding MSII, improvements are incorporated and a new and better segmentation algorithm called "MSII version 2" has been proposed.

Files in this item

Files Size Format
b14719666.pdf 4.302Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information