Digital Image Processing
`
`_
`
`1A1 1977
`
`Addison-Wesley Publishing Company
`Advanced Book Program
`Reading, Massachusetts
`
`London - Amsterdam - Don Mills, Ontario . Sydney . Tokyo
`
`ZEISS, et aI., EX. ‘I013
`
`)MPUTATlON
`
`arence Works
`
`hern California
`
`0 Ordinary Differential
`
`cs, 1 9 73
`
`,
`fluons’
`
`I 9 74
`
`»Bz0sczence, Medical
`
`BA
`L
`r, J 9 74
`
`., and SUEO UENO
`
`Direct, 1 9 75
`
`1 Lag7angz'ans:
`1 9 75
`
`'ysz's, J 9 76
`
`, J 9 77
`
`Rafael C. Gonzalez
`
`Department of Electrical Engineering
`University of Tennessee, Knoxville
`
`Paul wintz
`
`School of Electrical Engineering
`Purdue University, Lafayette, Indiana
`and
`
`Wintek Corporation, Lafayette, Indiana
`
`

`
`Library of Congress Cataloging in Publication Data
`
`
`Gonzalez, Rafael C
`Digital image processing.
`
` (Applied mathematics and computation ; no. 13)
`
`V
`Includes bibliographies and index.
`1.
`Image processing.
`I. Wintz, Paul A., joint
`author.
`II. Title.
`'
`
`TA1632.G66
`ISBN: 0-201-02596-5
`
`
`621.38’ 01,31,
`
`77-10317
`
`ISBN: 0-201-02597-3 pbk.
`
`Reproduced by Addison-Wesley Publishing Company, Inc., Advanced Book Program, Reading,
`Massachusetts, from camera-ready copy prepared under the supervision of the authors.
`
`American Mathematical Society (MOS) Subject Classification Scheme (1970): 68-00, 68A45,
`62-04, 65C99, 42-00, 42A68, 42A76, 92-04, 93-00, 94-04, 94A05, 94A10.
`
`Copyright © 1977 by Addison-Wesley Publishing Company, Inc.
`Published simultaneously in Canada.
`
`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, electronic, mechanical, photocopying, recording, or
`otherwise, without the prior written permission of the publisher, Addison-Wesley Publishing
`Company, Inc., Advanced Book Program, Reading, Massachusetts 01867, U.S.A.
`
`/.
`
`Manufactured in the United States of America
`
`ABCDE FGHIJ-HA- 7987
`
`
`
`
`
`

`
`ACKNOWLEDGMENTS
`
`
`
`
`
`0. Figures 2.9 and
`
`Spectrum, vol. 2,
`“The Fast Fourier
`
`E-12, no. 1, 1969,
`4.37 from T. G.
`
`ual Model,” Proc.
`.d IV courtesy of
`Jurtesy of J. L.
`)m H. C. Andrews
`
`rum, vol. 9, 1972,
`on of Turbulence-
`
`293—297. Figure
`s. Figure 5.5 from
`Processing Linear
`. 131-138. Figure
`st Squares Estima-
`Trans. Computers,
`:ourtesy of the Jet
`C. K. Chow and
`=ft Ventricle from
`
`72, pp. 388—4l0.
`California. Figure
`'. Wallace, Purdue
`:rsity of Southern
`.utomatic Analysis
`-223. Figure 7.38
`
`3
`
`
`
`A
`
`t
`
`1
`INTRODUCTION
`
`One picture is worth more
`than ten thousand words.
`Anonymous
`
`
`
`1.1 BACKGROUND
`
`
`
`_,
`[}
`
`,.
`
`I.
`}{
`'1
`.9;
`"
`;;j
`1
`
`
`
`
`
`image processing methods stems from two
`in digital
`Interest
`principal application areas: improvement of pictorial information for human
`interpretation, and processing of scene data for autonomous machine percep-
`tion. One of the first applications of image processing techniques in the first
`category was in improving digitized newspaper pictures sent by submarine
`cable between London and New York. Introduction of the Bartlane cable
`picture transmission system in the early l920’s reduced the time required to
`transport a picture across the Atlantic from more than a week to less than
`three hours. Pictures were coded for cable transmission and then reconstruc-
`ted at the receiving end by specialized printing equipment. ‘Figure 1.1 was
`transmitted in this way and reproduced on a telegraph printer fitted with
`type faces simulating a halftone pattern.
`' Some of the initial problems in improving the visual quality of
`these early digital pictures were related to the selection of printing proce-
`dures and the distribution of brightness levels. The printing method used to
`1 obtain Fig. 1.1 was abandoned toward the end of 1921 in favor of a tech-
`nique based on photographic reproduction made from tapes perforated at
`the telegraphreceiving terminal. Figure 1.2 shows an image obtained using
`this method. The improvements over Fig. 1.1 are evident, both in tonal
`quality and in resolution.
`-
`'
`The early Bartlane systems were capable of coding images in five
`distinct brightness levels. This capability was increased to fifteen levels in
`1929. Figure 1.3 is indicative of the type of image that could be obtained
`using the fifteen-tone equipment. During this period,
`the reproduction
`process was also improved considerably by the introduction of a system for
`Hans: c. Gonzalaz and Fuul wnnu, mgnan Irnaga Fracesing
`xsan o-201-02595-5; O-‘Z01-02597<!(nbk.)
`Copyright © 1977
`av Addison-Waslav Publishing Company, Inc., Advanced Book Prugram. All rlghu
`rasarv-d. Nu par! at Ihls nublluulnn mav bu rapmducad, stared In : rntrluval system, or transmitted, in
`any term or bv any means. ulncxranic, mechanical phatocapvlnu, recording, or amarwlsa, without the
`nrlor narrnlulon of mu publisher.
`
`1
`
`

`
`
`
`2
`
`1
`
`INTRODUCTION]
`
`developing a fflm plate via light beams which were modulated by the coded
`picture tape.
`
`;;.«..«»‘»a.mm.«.-
`
`1.1 BACKGROUND
`
`
`
`Figure 1.1. A digital picture produced in 1921
`from a coded tape by a telegraph printer with
`special type faces. (From McFarlane [1972]/.)
`
`Although improvements on processing methods for transmitted
`digital pictures continued to be made over the next thirty-five years, it took
`the combined advents of large-scale digital computers and the space program
`to bring into focus the potentials of image processing concepts. Work on
`using computer techniques for improving images from a space probe began at
`the Jet Propulsion Laboratory (Pasadena, California) in 1964, when pictures
`of the Moon transmitted by Ranger 7 were processed by a computer to cor-
`rect -various types of image distortion inherent in the on-board television
`
`camera. These techniques served as the basis for improved methods used in
`the enhancement and restoration of images from such familiar programs as
`the Surveyor missions to the Moon, the Mariner series of flyby missions to
`Mars, and the Apollo manned flights to the Moon.
`From 1964 until
`this writing,
`the field of image processing has
`experienced vigorous growth.
`In addition to applications in the space
`program, digital image processing techniques are used today in a variety of
`
`
`
`Figure 1.2. A digital picture made from a tape punched after the signals had crossed the
`Atlantic twice. Some errors are visible. (From McFarlane [1972] .)
`
`Figure 1.3. Unret
`tone equipment fr
`
`problems whic
`methods capabi
`tion and analy
`computer procc
`into color for -
`The same or sin
`
`patterns from 2
`tion procedure:
`coverable objec
`have been ins1
`which were the
`I being photogra
`methods. In pl
`as high-energy
`computer tech
`concepts can bu
`ment, defense,
`Some
`
`processing tecl
`on the left an:
`
`Figure l.4(a) i
`interference d1.‘
`which in this c
`
`completely 1'61
`Figures l.4(c)
`
`
`
`

`
`1.1 BACKGROUND
`
`
`
` RODUCTIONI
`
`the coded
`
`ced in 1921
`printer with
`ne [197 2] .)
`
`zransmitted
`
`tars, it took
`ce program
`;. Work on
`
`be began at
`L611 pictures
`uter to cor-
`1 television
`ods used in
`
`programs as
`missions to
`
`acessing has
`the space
`a variety of
`
`ad crossed the
`
`
`
`Figure 1.3. Unretouched cable picture of Generals Pershing and Foch, transmitted by 15-
`tone equipment from London to New York. (From McFarlane [1972] .)
`
`problems which, although often unrelated, share a common need for
`methods capable of enhancing pictorial information for human interpreta-
`tion and analysis.
`In medicine,
`for instance, physicians are assisted by
`computer procedures that enhance the contrast or code the intensity levels
`into color for easier interpretation of x-rays and other biomedical images.
`The same or similar techniques are used by geographers in studying pollution
`patterns from aerial and satellite imagery. Image enhancement and restora-
`tion procedures have been used to process degraded images depicting unre-
`coverable objects or experimental results too expensive to duplicate. There
`have been instances in archeology, for example, where blurred pictures
`which were the only available records of rare artifacts lost or damaged after
`"being photographed, have been successfully restored by image processing
`methods. In physics and related fields, images of experiments in such areas
`as high-energy plasmas and electron microscopy are routinely enhanced by
`computer techniques. Similar successful applications of image processing
`concepts can be found in astronomy, biology, nuclear medicine, law enforce-
`ment, defense, ahd industrial applications.-
`Some typical examples of the results obtainable with digital image
`processing techniques are shown in Fig. 1.4. The original images are shown
`on the left and the corresponding computer-processed images on the right.
`Figure 1.4(a) is a picture of the Martian surface which was corrupted by
`interference during transmission to Earth by a space probe. The interference,
`which in this case appears as a set of vertical, structured lines, can be almost
`completely removed by computer processing, as shown in Fig.
`l.4(b).
`Figures 1.4(c) and (d) illustrate the considerable improvement that can be
`
`

`
`
`
`INTRODUCTION
`
`1.2 DIGITAL IM
`
`made on an
`
`shown in Fig
`the image shr
`rithm. These
`—4and 5.
`
`The
`
`results are in
`
`area of digit
`this section
`interest is fo
`in a form st
`little resemb
`content of a’
`
`perception :
`multidimens:
`
`Ty:
`1mage proce
`
`robots for p
`tic processir
`machine pro
`crop assessm
`
`1.2 DIGITA
`
`As
`refers to a '
`
`denote spat
`tional
`to t]
`
`example I111.‘
`ters is show
`
`perspective
`this way it '
`changes in
`brightness 1
`of assigning
`the compo]
`brightness ii
`
`..A «
`
`_in spatial cc
`matrix who
`
`correspondi
`I The elemer
`
`elements,
`
`12
`
`
`
`
`
`Figure 1.4. Examples of computer-processed images.
`
`
`
`

`
`
`
`‘RODUCTION
`
`1.2 DIGITAL IMAGE REPRESENTATION
`
`-
`
`5
`
`
`
`
`
`made on an x-ray image by contrast and edge enhancement. The image
`shown in Fig.
`l.4(e) was blurred by uniform motion during exposure, and
`the image shown m Fig. l.4(f) resulted after application of a deblurring algo-
`rithm. These illustrations are typical of those discussed in detail
`in Chapters
`4 and 5.
`
`-
`
`
`
`The foregoing examples have in common the fact that processing
`results are intended for human interpretation. The second major application
`area of digital image processing techniques mentioned at the beginning of
`this section is in problems dealing with machine perception. In this case,
`interest is focused on procedures for extracting from an image information
`in a form suitable for computer processing. Often, this information bears
`little resemblance to visual features used by humans in interpreting the
`content of an image. Examples of the type of information used in machine
`perception are statistical moments, Fourier transform coefficients, and
`multidimensional distance measures.
`Typical problems in machine perception which routinely employ
`image processing techniques are automatic character recognition, industrial
`robots for product assembly and inspection, military recognizance, automa-
`tic processing of -fingerprints, screening of x-rays and bloodgpsamples, and
`machine processing of aerial and satellite imagery for weather prediction and
`
`crop assessment.
`’
`
`
`1.2 DIGITAL IMAGE REPRESENTATION
`
`As used in this book, the term monochrome image or simply image,
`refers to a two-dimensional light intensity function f(x, y), where x and y
`denote spatial coordinates and the value of f at any point (x, y) is propor-
`tional
`to the brightness (or gray level) of the image at that point. An
`example illustrating the axis ‘convention used throughout the following chap-
`ters is shown in Fig. 1.5. It is sometimes useful to view an image function in
`perspective with the third axis being brightness. If Fig. l.5 were viewed in
`this way it would appear _as=a series of active peaks in regions with numerous
`changes in brightness levels and smoother regions or plateaus where the
`brightness levels" varied little or were constant. If we follow the convention
`of assigning proportionately higher values to brighter areas, the height of
`the components in the plot would be proportional to the” corresponding
`brightness in the image.
`
`
`
`._
`.-
`-.'
`“A dz’
`_
`_ whih - _ -‘,1 .1.
`in spatial coordinates and in brightness. We may consider a digital image as a
`matrix whose row and column indices identify a point in the image and the
`corresponding matrix element value identifies the gray level at that point.
`The elements of- such a digital array are called image elements, picture
`elements, pixels, or pels, with the last two names being commonly used
`
`
`
`

`
`
`
`g
`
`6
`
`.
`
`1 INTRODUCTION
`
`1.3 ELEMEI
`
`abbreviations of “picture elements”.
`‘
`Although the size of a digital image varies with the application, it
`will become evident in the following chapters that there are numerous advan-
`tages in selecting square arrays with sizes and number of gray levels which
`are integer powers of 2. For example, a typical size comparable in quality to
`a monochrome TV image is a 512 X 512 array with 128 gray levels.
`
`‘
`
`I
`
`,———_ Origin
`
`1
`
`Image
`
`1.3.1 Digi
`
`A
`
`ble for in]
`devices at
`TV camer
`tized be i]
`
`Image dis:
`but they
`images th:
`I
`on a flat '
`
`ing a bear
`in relatio
`
`through 1
`image. In
`level at a
`
`1
`
` K
`
`
`
`Figure 1.5. Axis convention ‘used for digital image representation.
`
`With the exception of a discussion in Chapter, 4 of pseudo-color
`techniques for image enhancement, all the images considered in this book are
`digital monochrome images of the form described above‘;"_,Thus, we will not
`be concerned with topics in three-dimensional scene analysis nor with optical
`techniques for image processing.
`”
`_
`
`1.3 ELEMENTS OF A DIGITAL IMAGE PROCESSING SYSTEM
`
`The components of a basic, general-purpose digital image processing
`system are shown in Fig. l.6.'_The operation of such a system may be divided
`into three principal categories: digitization, processing, and display."
`
`
`
`

`
`)UCTION
`
`V:
`
`1.3 ELEMENTS OF A DIGITAL IMAGE PROCESSING SYSTEM
`
`Mass
`
`Storage
`
`Digital
`Computer
`
`Operator
`Console
`
`Figure 1.6. Elements of a digital image processing system.
`
`1.3.1 Digitizers /
`
`0-color
`30k are
`vill not
`
`A digitizer converts an image into a numerical representation suita-
`ble for input into a digital computer. Among the most commonly used input
`devices are microdensitometers, flying spot scanners, image dissectors, and
`TV camera digitizers. The first two devices require that the image to be digi-
`tized be in the form of a transparency '(e.g., a film negative) or photograph.
`Image dissectors and TV cameras can accept images recorded in this manner,
`but they have the additional advantage of being able. to digitize natural
`images that have sufficient light intensity to excite the detector.
`In microdensitometers the transparency or photograph is mounted
`on a flat bed or wrapped around a drum. Scanning is accomplished by focus-
`ing a beam of light on the image and translating the bed or rotating the drum
`in relation to the beam, In _the case of transparencies the beam passes
`through the film; in photographs it is reflected from the surface of the
`image. In both cases the beam is focused on a photodetectortand the gray
`level at any point in the image is recorded by the detector based on the
`
`messing
`divided
`
`V
`
`
`
`

`
`1
`
`INTRODUCTION
`
`1.3 ELEMENTS C
`
`intensity of the beam. A digital image is obtained by allowing only discrete
`values of intensity and position in the output. Although microdensitometers
`are slow devices, they are capable of high degrees of position accuracy due to
`the essentially- continuous nature of the mechanical translation used in the
`digitization process.
`
`Flying spot scanners also operate-on the principle of focusing a
`transmitted or reflected source beam on a photodetector. In this case, how-
`ever, the image is stationary and the light source is a cathode-ray tube (CRT)
`in which a beam of electrons, deflected by electromagnets, impinges on a
`fluorescent phosphor surface. The beam thereby produces a spot of light
`that moves in a scanning pattern on the face of the tube. The fact that the
`beam is moved electronically allows high scanning speeds. Flying spot scan-
`ners are also ideally suited for applications in which it is desirable to control
`the beam scanning pattern externally (e.g.,
`in tracing the boundaries of
`objects in an image). This flexibility is afforded by the fact that the position
`of the electron beam is quickly and easily established by external voltage
`signals applied to the electromagnets.
`In image dissectors and TV cameras the image is focused directly on
`the surface of a photosensitive tube whose response is proportional to the
`incident light pattern-. Dissector operation is based on the principle of elec-
`tronic emission, where the image incident on the photosensitive surface
`produces an electron beam whose cross section is roughly the same as the
`geometry of the tube surface. Image pickup is accomplished by using electro-
`magnets to deflect the entire beam past apinhole located in the back of the
`dissector tube. The pinhole lets through only a small cross section of the
`beam and thus “looks” at one point in the image at a time. Since photoemis-
`sive materials are very inefficient, the time that the pinhole has to look at
`the point source in order to collect enough electrons tends to make image
`dissectors rather slow digitizers. Most devices integrate the emission of each
`input point over a_-‘specified time interval before yielding a signal which is
`proportional
`to the brightness of the point. This integration capability is
`beneficial in terms of noise reduction, thus making image dissectors attrac-
`tive in applications where high signal-to-noise ratios are required. Like in
`flying spot scanners, control of the scanning pattern in image dissectors is
`easfly varied by external voltage signals applied to the electromagnets.
`
`
`
`Althc
`
`than the syste:
`many applicat.
`for example, :1
`also have the 1
`
`in its entirety
`
`systems discus
`con systems
`reasonably ef:
`meters and in
`beam control
`
`required to p1
`monitor.
`
`1.3.2 Image Pr
`
`Syste
`devices for spa
`performing a
`principal parai
`processing is 1
`applications V
`equipped min
`in most cases
`
`the key in st]
`bulk storage c
`
`I
`
`tapes and dis]
`device, depen-
`arrays. Generz
`used during p:
`computer met
`In tr
`
`computer sys
`computer wit]
`magnetic tape
`recorders,
`an
`Although ima
`to gain speed,
`having at lea:
`A sy‘
`difficulty to i
`the following
`the size of the
`
`I
`
`
`
`f
`l
`
`Many general-purpose TV image digitizers employ a vidicon tube,
`
`whose operation is based on the principle of photoconductivity. An image
`focused on the tube surface produces a pattern of varying conductivity
`which matches the distribution of brightness in the optical image. An inde-
`‘\_ pendent, finely focused electron beam scans the rear surface of the photo-
`'3 conductive target, and by charge neutralization, this beam creates a poten-
`lltial difference and produces on a collector a signal proportional to the input
`I.‘ brightness pattern. A digital image is obtained by quantizing this signal, as
`kwell as the corresponding position of the scanning beam.
`
`

`
` 1.3 ELEMENTS OF A DIGITAL IMAGE PROCESSING SYSTEM
`INTRODUCIION
`9
`
`
`
`\
`
`
`
`
`
`(\
`
`
`1.3.2 Image Processors
`
`
`lensitometers _
`
`:uracy due to
`1 used in the
`
`pf focusing a
`Lis case, how-
`1 tube (CRT)
`npinges on a
`spot of light
`fact that the
`
`ng spot scan-
`ale to control
`
`»ound_aIies of
`t the position
`tern al voltage
`
` only discrete
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Although standard Vidicon digitizers are in general less accurate
`than the systems discussed above, they have numerous advantages which in
`many applications outweigh their relative lack of precision. Vidicon "systems,
`for example, are among the most inexpensive digitizers on the market. They
`also have the distinct advantage that the image being digitized can be viewed
`in its entirety on a TV monitor. This capability, not available in any of the
`systems discussed above, is ideal for general-purpose applications. Since Vidi-
`con systems employ electronic scanning,‘ and photoconductive tubes are
`reasonably efficient,
`these digitizers are much faster than microdensito-
`meters and image dissectors. They are not as fast or as flexible in terms of
`/’
`beam control as flying spot scanners because of the scanning constraints
`required to produce a video image which can be viewed"on a standard TV ‘i
`monitor.
`
`
`
`Systems used for image processing range from microprocessor
`devices for special-purpose applications to large cornputersystems capable of
`performing a variety of functions on high-resolution image arrays. The
`
`principal parameter influencing the structure of a computer system forimage
`
`processing is the required data throughput. For general-purpose laboratory
`
`applications where fast data‘ throughput
`is not essential, a moderately
`
`
`equipped minicomputer is often adequate. Since digitized image arrays are
`in most cases too large to fit completely in the memory of a small computer,
`
`
`the key in structuring such a system is to provideadequate and efficient
`
`bulk storage capabilities. The two most popular storage media are magneti
`
`tapes and disk packs, both of which allow storage of numerous images per
`
`
`device, depending on the density of the medium and the size of the image
`arrays. Generally, magnetic tapes are used for archival storage, and disks are
`
`
`used during processing in order to improve data transfer speed between the
`
`computer memory and bulk storage media.
`flexible
`In terms of the above requirements a minimum, yet
`
`computer system for image processing can be structured using a mini-
`
`
`computer with 32,000 to 64,000 words of core memory, two disk drives, a
`magnetic tape unit, and assorted peripherals such as scope terminals, cassette
`
`
`recorders, and a lineprinter or some other hard-copy output device.
`
`Although image processing programs are often coded in assembly language
`to gain speed, the flexibility of the system can be improved considerably‘ by
`
`having at least one high-level
`language for use. in program development.
`
`A system with the components just described can be used without
`
`
`difficulty to implement most of the image processing methods discussed in
`
`the following chapters. The operating efficiency will, of course, depend on
`g this signal, as
`the" size of the input images and the type of processing desired.
`
`
`ad directly on
`rtional to the
`
`lciple of elec-
`sitive surface
`e same as the
`
`using electro-
`1e back of the
`section of the
`
`ce photoemis-
`has to look at
`0 make image
`Iission of each
`
`iignal which is
`1 capability is
`sectors attrac-
`uired. Like in
`
`ge dissectors is
`lectromagnets.
`L Vidicon tube,
`
`vity. An image
`g conductivity
`nage. An inde-
`= of the photo-
`reates a poten-
`ial to the input
`
`-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`

`
`
`
`10
`
`1
`
`INTRODUCTION
`
`1.4 ORGANIZI
`
`these voltage
`In both of t]
`ters constitu
`to drive the '
`Prir
`
`resolution in
`
`tone images
`dard line-pr
`controlled b
`
`point. By pr
`able good gr
`
`tively few cl
`Other comrr
`
`sensitive pap
`
`1.4 ORGAN
`
`Tec
`
`(
`categories:
`(3) image e
`
`material in 1
`as these pro]
`As
`
`of convertir
`
`components
`quality is g
`sampling prt
`natural exte
`introductior
`the book.
`En]
`
`ment of a g
`ment is the
`
`in Chapter f
`reduce the
`
`plays a cent:
`or transmiss
`are consider
`
`position of
`parts in a :
`consideratio
`
`recognition
`
`1.3.3 Display Devices
`
`The function of the display unit in an image processing system is to
`convert the numerical arrays stored in the computer into a form suitable for
`human interpretation. The principal display media are CRTs, TV systems,
`and printing devices.
`‘
`In CRT systems the horizontal and vertical positions of each
`element in the image array are converted into voltages which are used to
`deflect the CRT’s electron beam, thus providing the two-dimensional drive
`necessary to produce an output image. At each deflection point, the inten-
`sity -of the beam is modulated by using a voltage which is proportional to the
`value of the corresponding point in the numerical array, varying from zero
`intensity outputs for points whose numerical value corresponds to black, to
`maximum intensity for white points. The resulting variable-intensity light
`pattern is recorded by a photographic camera focused on the face of the
`cathode-ray tube. Some systems employ a long persistence phosphor tube
`which also allows viewing of the entire image after the scanning process is
`completed. Although images recorded by the photographic process can be of
`excellent quality, the same images generally appear of poor tonality when
`shown to an observer on a long persistence CRT because of limitations in the
`human visual system when responding to this type of display.
`Television display systems convert an image stored in the computer
`f
`into a video frame which can be displayed on a TV monitor.-The advantage
`1
`of these systems is that displays created on a video monitor have a tonality
`which closely resembles that of photographs, thus producing an output
`' which is easily assimilated by the visual system. The disadvantage of TV
`displays is that they must refresh the monitor at a rate of about 30 frames
`(e.g.,
`images) per second in order to avoid flicker. Since most general-
`purpose computers are not capable of transferring data at this rate,‘ the
`principal problem in designing a TV display system is to provide some buffer
`Zfitorage medium for transferring data to a’ monitor at video rates. Most high-
`quality commercial systems accomplish this in one’ "of ‘two ways. One
`approach is to use a fast solid-state memory to store the entire image array;
`the screen is then refreshed at 30 frames per second by cycling through the
`memory and combining the stored binary information into an analog signal
`by means of conditioning circuits and fast digital-to-analog converters. The
`other method is to store the image array on a high-density disk. The storage
`' arrangement is again in binary form, with each track in the disk containing a
`bit for all the pixels in the image. An N X N image with 2" possible levels,
`for example, requires 11 storage tracks, each-containing’ N X N bits. The
`necessary transfer speed is accomplished by rotating the disk past n sensing
`heads. For any disk position,
`the binary information in all
`71 heads is
`combined to produce a voltage proportional to the gray level of a single pixel
`in the image. As the disk rotates, an analog signal is created by conditioning
`
`I ‘
`
`.i
`
`
`
`

`
`ITRODUCTION
`
`1.4 ORGANIZATION OF THE BOOK
`
`
`
`
`
`these voltage levels and inputing them into a fast digital-to-analog converter.
`In both of the approaches just discussed the analog signals out of the conver-
`ters constitute the information carrying component of the video signal used
`to drive the TV display monitor.
`low-
`Printing image display devices are useful primarily for
`resolution image processing work. One simple approach for generating gray-
`tone images directly on paper is to use the over-strike capability of a stan-
`dard line-printer. The gray-level of any point
`in the printout can be
`controlled by the number and density of the characters overprinted at that
`point. By properly selecting the character set it is possible to achieve reason-
`able good gray-level distributions with a simple computer program and rela-
`tively few characters. An example of this approach is given in Appendix A.
`Other common means of recording an image directly on paper include heat
`sensitive paper devices and ink spray systems.
`
`
`
`system is to
`suitable for
`
`FV systems,
`
`ns of each
`are used to
`lsional drive
`
`t, the inten-
`iional to the
`
`g from zero
`to black, to
`tensity light
`face of the
`
`osphor tube
`1g process is
`ass can be of
`»nality when
`ations in the
`
`he computer
`1e advantage
`ve a tonality
`; an output
`ntage of TV
`it 30 frames
`
`lost general-
`his rate, the
`some buffer
`
`s. Most high-
`- ways. One
`image array;
`5 through the
`analog signal
`iverters. The
`. The storage
`containing a
`)ssible levels,
`N bits. The
`
`-ast n sensing
`l n heads is
`
`a single pixel
`conditioning
`
`1.4 ORGANIZATION OF THE BOOK
`
`Techniques for image processing may be divided into four principal
`categories: (1) image digitization; (2) image enhancement and restoration;
`(3) image encoding; and (4) image segmentation and representation. The
`material in the following chapters is organized in essentially the same order
`as these problem areas.
`As discussed in Sections 1.2 and 1.3, the digitization problem is one
`of converting continuous brightness and spatial coordinates into discrete
`components. A preliminary discussion of digitization and its effect on image
`quality is given in Chapter 2, while a more theoretical treatment of the
`sampling process is developed in Chapter 3. Digitization considerations are a
`natural extension of the main theme of these two chapters, which is the
`introduction of concepts and mathematical tools used throughout the rest of
`the book.
`Enhancement and restoration techniques deal with the improve-
`ment of a given image for human or machine perception. Image enhance-
`ment is the topic of Chapter 4, while image restoration methods are covered
`in Chapter 5. Image encoding procedures,_discussed in Chapter 6, are used to
`reduce the number of bits in a digital image. The encoding process often
`plays a central role in image processing for the purpose of minimizing storage
`or transmission requirements. Segmentation and representation techniques
`are considered in Chapter 7. These procedures, which deal"with the decom-
`position of an image into a set of simpler parts and the organization of these
`parts in a meaningful descriptive manner, are among the most important
`considerations in the development of autonomous image processing and
`recognition systems.
`
`.
`
`
`
`

`
`
`
`Complementary reading for the material in this book may be found in
`the books by,Rosenfeld [I969], Andrews [l970], Lipkin and Rosenfeld
`[1970], Duda and Hart [1973] , Huang [I975], Andrews and Hunt [1976] ,
`and Rosenfeld and Kak [1976] . The following special issues of journals have
`been devoted to image processing: Proc.
`IEEE [l972], IEEE Trans.
`Computers [I972], Computer [l974], and IEEE Trans. Circuits Syst.
`[I975]. Other survey articles of interest are by Andrews, Tescher, and
`Kruger [I972], Rosenfeld [l972, 1973, 1974], Andrews [I974], and Fu
`and Rosenfeld [I976]. The pattern recognition literature often contains
`articles related to image processing. The books by Tou and Gonzalez
`[I974], and Gonzalez and Thomason (in preparation) contain a guide to
`the literature on pattern recognition and related topics.
`
`1
`
`INTRODUCTION
`
`REFERENCES
`
`The references cited below are of a general nature and cover the
`spectrum of available image processing techniques and their applications.
`References given at the end of later chapters are keyed to specific topics
`discussed in the text. All references are cited by author, book, or journal
`name followed by the year of publication. The bibliography at the end of
`the book is organized in _the same way and contains all pertinent information
`for each reference.
`
`The 1
`of image cc
`throughout
`of the hum
`
`,
`
`capabilities
`sents an in
`
`which gives
`concepts of
`duced in Se
`Section 2.4
`
`ly, Section ’.
`most impor
`
`2.1 ELEME
`
`Since
`
`following ci
`image, it is
`the visual‘—p-
`the human
`serve as f0‘
`
`2.1.1 Struc
`
`A h(
`2.1. The e
`Hafall C. Gunzalaz and.
`cnpvrlghz ©1971 by
`rasarvud. No part af 1|
`any form or by any n
`prior pumissian af mo