Degree Type

Dissertation

Date of Award

1987

Degree Name

Doctor of Philosophy

Department

Electrical and Computer Engineering

First Advisor

Arthur V. Pohm

Abstract

To investigate different codes that are to be used in compressing data in an electronic library system, the author developed an image data base, at the resolution of the IBM PC (640 x 200 x 2), that represents typical library materials. The CCITT standard one and two dimensional compression techniques were applied to this data base and gave a compression factor (c.f.) less than the c.f. reported in the literature at higher resolution. The standards gave a very low c.f. for images that contain a lot of text. The Lempel-Ziv-Welch (LZW) algorithm, which is usually used to compress text, was tried on the images. LZW gave better c.f. for text only images and showed promising results for other types of images. The decompression time (d.t.) of LZW is much smaller than that of the CCITT code;Three different ways to optimize LZW for compression of screens were investigated. The optimization was carried by compressing the run-lengths of the screen pels. These methods showed a small improvement in the c.f. compared to LZW at much smaller d.t;Three different versions of modifying LZW to recognize longer strings in the input were investigated and gave an improvement in the c.f. Two out of these three versions were originally proposed by the author and gave slightly less c.f. than the third one, but at much smaller compression time (c.t.). The c.t. of the third method may restrict its use to nonreal time compression, such as the case of an electronic library system;The author proposed two methods to scan a screen or read a scanned screen, that divide the screen into horizontal blocks and scan each block column by column. These methods resulted in an increase in the c.f. that is specially obvious in the case of textual screens where the c.f. is up to 4.48 times higher than the c.f. of the CCITT standards. The c.f. is so high that these methods offer a better and simpler alternative to pattern recognition in compressing the images.

DOI

https://doi.org/10.31274/rtd-180813-8654

Publisher

Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/

Copyright Owner

Mansour Alsulaiman

Language

en

Proquest ID

AAI8805038

File Format

application/pdf

File Size

448 pages

Share

COinS