IMAGE AND TEXT COMPRESSION

IMAGE AND TEXT COMPRESSION

THE KLUWER INTERNATIONAL SERIES IN ENGINEERING AND COMPUTER SCIENCE COMMUNICATIONS AND INFORMATION TIlEORY Other books in the series: Consulting Editor: Robert Gallager Digital Communication. Edward A. Lee, David G. Messerschmitt ISBN: 0-89838-274-2 An Introduction to Cryptology. Henk C.A. van Tilborg ISBN: 0-89838-271-8 Finite Fields for Computer Scientists and Engineers. Robert 1. McEliece ISBN: 0-89838-191-6 An Introduction to Error Correcting Codes With Applications. Scott A. Vanstone and Paul C. van Dorschot ISBN: 0-7923-9017-2 Source Codi~g Theory. Robert M. Gray ISBN: 0-7923-9048-2 Switching and Traffic Theory for Integrated Broadband Networks. Joseph Y. Hui ISBN: 0-7923-9061-X Advances in Speech Coding, Bishnu Atal, Vladimir Cuperman and Allen Gersho ISBN: 0-7923-9091-1 Coding: An Algorithmic Approach, John B. Anderson and Seshadri Mohan ISBN: 0-7923-9210-8 Third Generation Wireless Information Networks, edited by Sanjiv Nanda and David J. Goodman ISBN: 0-7923-9128-3 Vector Quantization and Signal Compression, by Allen Gersho and Robert K1. Gray ISBN: 0-7923-9181-0

IMAGE AND TEXT COMPRESSION edited by James A. Storer Brandeis University ~. " SPRINGER SCIENCE+BUSINESS MEDIA, LLC

Libral'Y of Congress Cataloging.in.Publication Data Image and text compression / edited by James A. Storer. p. cm. -. (Kluwer international series in engineering and computer science. Communications and information theory) Includes bibliographical references and index. ISBN 978-1-4613-6598-3 ISBN 978-1-4615-3596-6 (ebook) DOl 10.1007/978-1-4615-3596-6 1. Data compression (Telecommunication) 2. Text processing (Computer science) 3. Image processing- Digital techniques. I. Storer, James A. (James Andrew). 1953- II. Series. TK5102. 5.148 1992 005.74' 6--dc20 92-17339 CIP Copydght 1992 by Springer Science+Business Media New York Originally published by Kluwer Academic Publishers in 1992 Softcover reprint of the hardcover 1st edition 1992 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, mechanical, photo-copying, recording, or otherwise, without the prior written permission of the publisher, Springer Science+Business Media, LLC. Primed 011 acid-free paper. This printing is a digital duplication of the original edition.

CONTENTS INTRODUCTION vii I. Part 1: IMAGE COMPRESSION 1 1. Image Compression and Tree-Structured Vector Quantization 3 R. M. Gray, P. C. Cosman, E. A. Riskin 1. Fractal Image Compression Using Iterated Transforms Y. Fisher, E. W. Jacobs, R. D. Boss 3S 3. Optical Techniques for Image Compression J. H. Reif, A. Yoshida n. Part 1: TEXT COMPRESSION 83 4. Practical Implementations of Aritlunetic Coding P. G. Howard, J. S. Vitter 8S S. Context Modeling for Text Compression 113 D. S. Hirschberg, D. A. Lelewer 6. Ziv-Lempel Compressors with Deferred-Innovation M. Cohn 14S

vi m. 7. Massively Parallel Systolic Algorithms for Real-Time Dictionary-Based Text Compression J. A. Storer Part 3: CODING THEORY 159 179 8. 9. 10. Variations on a Theme by Gallager R. M. Capocelli, A. De Santis On the Coding Delay of a General Coder M. J. Weinberger, A. Lempel, J. Ziv Finite State Two-Dimensional Compressibility D. Sheinwald 181 215 253 BmLIOGRAPHY INDEX 277 351

Introduction James A. Storer Computer Science Dept. Brandeis University Waltham, MA 02254 Data compression is the process of encoding a body of data to reduce storage requirements. With Lossless compression, data can be decompressed to be identical to the original, whereas with lossy compression, decompressed data may be an acceptable approximation (according to some fidelity criterion) to the original. For example, with digitized video, it may only be necessary that the decompressed video look as good as the original to the human eye. The two primary functions of data compression are: Storage: The capacity of a storage device can be effectively increased with data compression software or hardware that compresses a body of data on its way to the storage device and decompress it when it is retrieved. Communications: The bandwidth of a digital communication link can be effectively increased by compressing data at the sending end and decompressing data at the receiving end. Here it can be crucial that compression and decompression can be performed in real time. Key types of data to which compression technology is currently being applied include natural language text, computer source and object code, bit-maps, numerical data, graphics, CAD data, map and terrain data, speech, music, scientific and instrument data, fax / half-tone data, gray-scale or color images, medical data and imagery, video, animation, and space data.

viii An important issue with all types of data is how much speed is required to process the data in real time; the following list gives an idea of the diverse range of speeds for different practical applications: Text sent over a modem'" 2,400 bits per second (Depending on the cost of the modem, commonly used speeds range from 1,200 bits per second to 9,600 bits per second.) Speech'" 100,000 bits per second (One government standard uses 8,000 samples per second, 12 bits per sample.) Stereo Music'" 1.5 million bits per second (A standard compact disc uses 44,100 samples per second, 16 bits per sample, 2 channels.) Picture Phone", 12 million bits per second (A low resolution black and white product might require 8 bits per pixel, 256x256 pixels per frame, 24 frames per second.) Black&White Video'" 60 million bits per second (A medium resolution product might use 8 bits per pixel, 512 by 512 pixels per frame, 30 frames per second.) HDTV '" 1 billion bits per second (A proposed standards has 24 bits per pixel, 1024 by 768 pixels per frame, 60 frames per second.) Although complete compression systems often employ both loss less and lossy methods, the techniques used are typically quite different. The first part of this book addresses lossy image compression and the second part lossless text compression. The third part addresses techniques from coding theory, which are applicable to both lossless and lossy compression. The chapters of this book were contributed by members of the program committee of the First Annual IEEE Data Compression Conference, which was held in Snowbird Utah, April, 1991.