A Colorimetric performance comparison of gray component replacement algorithms

Theses

Thesis/Dissertation Collections

5-1-1998

A Colorimetric performance comparison of gray

component replacement algorithms

Thomas Orino

Follow this and additional works at:

http://scholarworks.rit.edu/theses

This Thesis is brought to you for free and open access by the Thesis/Dissertation Collections at RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please [email protected].

Recommended Citation

A

Colorimetric

Performance Comparison

of

Gray

Component

_{ReplacementAlgorithms}

by

ThomasP. Orino

Athesissubmitted in partialfulfillmentofthe

requirements forthe degree ofMaster ofScience in the Schoolof

Printing

Managementand Sciencesin the

Collegeof

Imaging

Artsand Sciencesof the Rochester Institute of

Technology

May,

1998

Certificate of Approval

Master's Thesis

This is to certify that the Master's Thesis of

Thomas

P.

Orino

With a major in Printing Technology

has been approved by the Thesis Committee as satisfactory

for the thesis requirement for the Master of Science degree

at the convocation of

May, 1998

Date

Thesis Committee:

Steve Viggiano

Thesis Ad visor

Joseph

L.

Noga

Graduate Program Coordinator

C.

Harold Goffin

PermissiontoReproduce Thesis

Colorimetric Performance Comparisonofthe

Gray

ComponentReplacement

Functions Used in High-End and

Desktop

Separations

I,

ThomasP.

Orino,

hereby

grantpermissionto theWallace Memorial

Library

of

R.I.T. to reproduce this thesisinwholeorinpart.

Any

reproductionwill notbe for commercial use orprofit.

Chapter Page

List ofFigures iv

List ofTables v

Abstract vi

One Introduction 1

Endnotes forChapterOne 3

Two Theoretical Background 4

SubtractiveColor

Theory

4

GCR: Simple Model 6

GCR: Real Model 8

Benefits ofGCR 11

Color Measurement 13

Endnotes forChapterTwo 17

Three Review ofLiterature 19

Four Hypotheses 21

Five

Methodology

23

Endnotes forChapterFive 29

Six Results 30

Affect ofGCRLevel 37

Desktop

Algorithms vsHell ScannerSoftware 38

Comparison of

Desktop

Systems 39

The InfluenceofUnderColorAddition 40

Endnotes forChapterSix 42

Seven

Summary

andConclusions 43

Recommendations for Further

Study

44

Bibliography

45

Appendix A Original Data 49

Appendix B Calculated ColorDifferences 62

AppendixC Results ofANOVAAnalyses 71

ListofFigures

Figure Page

1 Spectral reflectance curvesforideal and

actual processinks 5

2

Gray

component of3-color overprint

(simple model) 6

3 Resultof50% GCR Separation (simplemodel) 6

4 Resultof 100% GCRseparation(simplemodel) 7

5 Near neutral testpatch 7

6

Gray

component oftestpatch (simple_model) 9

7 Actual _graycomponentdetermined

by

gray

balancetest 9

8 Resultof simple GCRseparation 10

9 Resultof GCRseparation

taking

gray

balance intoconsideration 10

10 RepresentationofCIELAB's

color space 14

11 Kodak Q60A Color Scanner Target 24

12 AreasoftheQ60Atarget measured and

analyzed inthis experiment ₂₇

1

Summary

statistics of the fourseparation methods

at 50% GCR 31

2

Summary

statistics ofthe fourseparationmethods

at80% GCR 31

3 Descriptive statistics ofdata takenfrom separations

performed at50% GCR 35

4 Descriptive statistics ofdata taken fromseparations

performed at 80% GCR 36

5

Summary

statistics_comparingthe fourseparation

methods at50% and 80% GCR 37

6

Summary

statistics

comparing

the

desktop

separation

methods withthe Hell scanner separations 39

7

Summary

statistics

comparing

the separations

performed

by

the two

desktop

methods 40

8

Summary

statistics

comparing

the separations performed

ontheHell scanner with and withoutUCA 41

Abstract

Anexperiment was performed tocomparethe effect of_graycomponent

replacementoncolor separations created withthe Hell 399ERscanner software compared to two

desktop

algorithms. Three different GCRmethods were _tested; the scanner method wheretheseparations were performed and filmsoutput on

a mid 1980'smodelHell 399ER laserscanner_{using Hell's first}generation GCR algorithmandtwo types of

desktop

methods where thescans weredonein

RGB,

color separated onthe

desktop,

andfilms generated onanAgfa Selectset 5000 imagesetter. Thetwo

desktop

methods used were RIT Research

Corporation's RGB-CMYKtransformandAdobe

Photoshop

2.0. All scans and

proofs were madeintheColorSeparation Labatthe Rochester Instituteof

Technology

(RIT)

in

Rochester,

New York. The

desktop

separations were made and filmoutputinRIT's Electronic Prepress Lab.

Foreachseparation_{method, three}levels ofGCRwere_performed; a non-GCR

(0%)

separation, one at50%

GCR,

and one at 80% GCR. In addition, the

separations performed ontheHell ScannerweredonewithUnder Color Addition

(UCA)

off and on.

Afterseparation, thescans were outputto film and proofed_using 3M Matchprint II. Twoproofs were made. Onecontained thesixseparations

spectrophotometer and theresults compared

using

the L*a*b*color space. All

color comparisons were done

using

the non-GCR separation as thecolor

reference. EachAE* measurement represents the colordifference betweena

patch on thenon-GCR targetand the

corresponding

patch onthe target

produced _{using GCR.}

Theresult oftheexperiment was the rejection of thestated hypotheses

that there would beno significant color difference betweenthe output produced

fromthe scanner separations and the

desktop

separations attwolevelsofGCR.

Theexperiment showed that the algorithms used to performGCR onthe

desktop

produced lesscolor variationthan the algorithm used in the Hell 399ER

scanner atboth50% and 80% GCR. The results also showed thatin almost

every case, the amount of color variationincreased as thelevel ofGCRwas

raised. Itcould notbe determined whether or notUnderColor Addition

(UCA)

had anysignificant influenceon color variationforthe separations performed

on theHell scanner.

Based on theresults of_{this experiment,} color professionals_usinga

desktop

productionworkflow shouldbeencouraged to takeadvantage ofthe

benefitsofGCRwithout fearthat the colorofthereproductions willbe

compromised.

Chapter One

Introduction

The 1988 SWOP handbook _{defines gray}component replacement

(GCR)

as "...a technique_{for removing from} the colorseparations some or all of the

cyan, magenta, and yellow thatproduces the

gray

component of a picture.The

gray

darkening

amounts are replaced

by

increasing

theblack printercontentin thesame area.

Simply

stated, GCRusesblackto create most oftheimage

shape and detail."1 GCR has great value to the_printing

industry

by

reducing

many

ofthe problems associated withthe multiple

layering

ofink on paper. As theprepress

industry

movestoward thelessexpensive and less proprietary

world ofthe

desktop,

itwillbe

necessary

for

desktop

hardware and softwareto produce _quality comparableto their high-end competition.

The

theory

behind

gray

ComponentReplacement

(GCR)

hasbeenaround

forover50years,2

butuntil

only

recently, the

technology

required toperform

theprocedure has been lacking. Mostoftoday's high-end digital scannershave

GCRalgorithmsbuilt

in,

and the

technology

isnow available in

desktop

prepress software. With the

ability

toproduce GCRseparations at thefingertips

of anyone who purchasesimage manipulation_software, it isimportantto

examinethe effectiveness ofGCR separationsperformed on

desktop

as well as

separations. Anexperiment was performed to determinethe color

differences

produced at

varying

levels ofGCR for both

desktop

and high-end electronic prepress paths. Since GCR is an alternativeto traditional electronic color

separation, it was importantto determine whether or not separations which utilize GCRproduce significant differences intheprinted result than traditional

color separations. For a given area onan _{original, there}shouldbe no

perceptiblecolor differencebetweena final output produced withtraditional

versusGCRseparations. Since thereis inherent_variabilityinthe

printing

process, no reproduction willbe a perfect color matchto the original,butit is importantto knowifthemethod of separation_undulyinfluences that

variability.

GCR isa

theoretically

sound_procedure,butif thealgorithms used to

performit are_{substandard, then the} separator cannot_adequatelyutilize

it,

and

Endnotes for Chapter

One

1 SWOP

Handbook,

p.

17,

1988.

2 John

Yule,

"FourColorProcesses and the Black Printer,"Journal

Theoretical Background

Subtractive Color

Theory

The basis ofcolor_printingis the reproductionofanoriginal

by

the

layering

of colored inks on a substrate.

Traditionally

this has been

accomplished withthe use ofthreeprocess inksand a blackskeletonprinter.

Theprocessinks are the threesubtractive primaries-cyan

(c),

magenta

(m),

and

yellow (y).

According

to subtractivecolor_theory,these three inksshould each

completely absorbone-third and reflectthe _remainingtwo-thirdsofthevisible

spectrum. The cyaninkshould absorb allincidentred

light,

the magenta ink

should absorb all incidentgreen

light,

and the yellow inkshould absorb all

incidentbluelight. Inorderto produce a desired _color, one would overprint

twoor three ofthe primaries.For_example, a red would beproduced

by

overprintingmagenta and yellowinks. Black wouldbe produced

by

a solid

overprintof all threeprimaries. In_{reality, the} light

absorbing

characteristics of

thepigments used inthe

formulation

of

printing

inksdo not

perfectly

matchthe

theoreticalmodel. Fig. 1 shows spectral reflectance curvesfor both

theoretically

idealinks and sample processinks. Because ofthe unwanted

// //

Ideal

yellow

400 500 600

Wavelength (nm)

Ideal

magenta

700

400 500 600

Typical

magenta

700

Wavelength (nm)

400

Ideal

cyan

Typical cyan

500 600 700

Wavelength

(nm)

Fig. 1 - Spectral

reflectance curves for

"ideal"

[image:14.569.152.418.75.606.2]

muddy

ofblack inkinto the

printing

process. Black ink isused toincreasethe tonal

range of the reproduction and to ensure thata neutralblackis produced in the

shadows.

Theaddition of theblackprinter to thereproduction process led to a

separationtechnique called undercolor removal (UCR). UCRwas developedto

make room fortheblack printerin theneutral shadow areas. Itisthe reduction

indot size oftheprocess colors where the blackprints and is used to

compensate for the addition of a blackink layerin the neutral shadows.

GCR : SimpleModel

An important aspect of process color

theory

is thata _graying

component is producedwhen the threeprimaries are overprinted.

Further,

with

CMYK Fig.2

-Gray

componentof3 color overprint(simplemodel). [image:15.569.343.497.439.640.2]

the ideal inks

described

previously, whenthree colors with thesame dotarea

are _{overprinted, the} result isa neutral (inactuality, this isnot the case and will

beexplained later). The size ofthedot determines the lightness ofthe neutral.

Forexample, the overprint of50% cyan, magenta, and yellowdots should pro

duce a neutral

gray

tint. In_theory,this sametint_mayalsobe created

by

printing

just a 50% blacktint. This

fact

lead to the observation

by

JohnYulein 1940

that,

"ifsuitable corrected negatives couldbemade easily, thebest results would usu

ally

be obtained

by

using

themaximum _quantity of

black,

and _printingnot

more than twoofthe threesubtractivecolors at_anyone point. A

brown,

for

instance,

wouldbe rendered

by

_magenta, yellow,and black insuitable propor

tions." *

According

to this theory,when_any three colors are _{overprinted, the}

colorwith thesmallestdot_size,combined with equal portions ofthe othertwo

colors, is the_graying component ofthe overprint.This is demonstrated in Fig. 2.

CMYK

Fig. 4

-Resultof100% GCR

separation (simplemodel).

CMYK

Fig.5

[image:16.569.344.497.434.633.2] [image:16.569.77.232.436.637.2]

nique, some or all of the

gray

componentis replaced withblackink. The

amount ofGCR canbevaried

depending

on the requirements ofthe reproduc

tion,

and isexpressed as a percentage ofthe _gray component replaced with

blackink. Fig. 3 and Fig.4 demonstrateGCRseparations of50and 100%.

It

may

benoted that ina 1986 study,

Gary

Field determined thatathigh

levelsof

GCR,

color variations increased.2 SWOPrecommends theuse ofGCR

at levels of50 to 80%.These levels ensure that theprinter receives thebenefits of

GCRwhile

avoiding

the increased _variabilityproduced athighlevels ofGCR.

Also,

at 100% GCRthe solid ink

density

intheshadows_{may fall} off as four ink

layers havebeenreplaced

by

only

one blackink layer. Undercoloraddition

(UCA)

may be

necessary

torestore acceptable shadow densities.3

GCR : Real Model

Aswas discussedearlier, theinksused in

printing

donotmatchthe ideal

theoretical inks described insimple subtractive color

theory

(seeFig.l).

Currently,

all process inkshavesome unwanted absorption characteristicsin

some part oftheirspectralcurves. This fact has greatimportance when onetries

to formulatean algorithm for

producing

effectiveGCRseparations.

Inthe simple theoretical example (Fig.

2),

the

graying

componentis pro

duced whenequal dotsizes of all threecolors are overprinted. In_{reality, the}

unwanted absorptions ofthe process inks throw this

gray balance

offsome

what. For _any giveninkset,

gray

balancetests must

be

performed inorder to

neutral. Formost process

inks,

the cyandotmustbe larger than the yellow and

magenta dots. Thisphenomenon must not be overlooked ifGCR is tobe

implemented correctly.4

The

following

example should

help

to illustratethedangerof

using

a

simpleGCRmodel ina near neutral area of a reproduction. Fig. 5 (page

7)

represents a near neutral patch

consisting

of a33% cyan, a 30%yellow

dot,

and

a30% magenta dot.A

gray

balance testperformed fortheseparticular_printing

conditionshas shown thatfor a33% cyan dot area to produce a _{neutral, the}

corresponding

magenta and yellowdotareas would be27%.

Therefore,

the _gray

componentis

33%C,

27%M,

and 27%Y (Fig.

7)

withthehueinformation carried

inthe 3%Mand 3%Y. 5 From this

information,

it isascertained that thispatchis

nearneutral witha

slightly

reddish cast. If thiswere separated _usingthe simple

model at 100%

GCR,

the 30% portionof all threecolors would be replaced with

a30% black tint(Fig. 6).

CMYK

Fig.6

-Gray

componentof

test patch (simplemodel).

CMYK

Fig. 7

[image:18.569.344.498.448.653.2]

duce a near neutral with a slight cyan cast.

Fig. 9 represents theresult ofthe GCR separationif the

gray balance

information isconsidered. Itproduces a near neutral with a reddishcast, which

was the expected result. The simple model's failureto consider the impurities

inthe inksprevents properreproduction of neutral areas.

Consequently,

it is

imperativethat

any

GCRmodel includesthe _{necessary gray balance}

informationforthe

printing

conditionsto beused inthe final reproduction.6

CMYK

Fig.8- Result_{of simple} GCRseparation.

CMYK

Fig.9

-ResultofGCRseparation_taking

[image:19.569.318.489.352.564.2] [image:19.569.78.232.356.563.2]

11

The Benefits

of

GCR

Gray

component replacementis an importantissue in the

printing

industry

because

of thebenefits itis capable of

providing

tothe printer. Some

of theclaimed

benefits

were scrutinized in Dr. Abdel

Ghaney

Saleh's 1984_paper, "Investigation into theApplication ofAchromatic Synthesis to the

Printing

Industry."7

ThemostimportantbenefitofGCR discussed isthe reductionin

sensitivity

tocolor

inking

fluctuations. When GCR isused, the neutral tones are produced with

mostly

black

ink,

_therefore, thereshouldbe littleorno color

fluctuationinthe

gray

tones. Thisisa much more stable conditionthan in

traditional

printing

where neutraltones are produced withthe overprint ofthe

three processinks. Intraditional printing, inkfluctuationsor changes indot

size (gainorsharpening)

may

upsetthe _{gray balance}

leading

tovisible and

unwanted hue shifts.

Therefore,

it is imperativethat theink filmthicknessis

carefully controlled overthe entire press run.

Since,

with

GCR,

theneutraltones are producedwith_mostlyblack

ink,

a change intheblack inkfilm thicknessor intheblackdotarea affects

only

thelightness. Thereis noshiftin hue and _gray balance ismaintained muchmore easily. This is a crucial point as _onlya slight

hue shiftin a neutral is_very noticeableto an observer. Adirect benefit of this

is,

with

GCR,

the press_{may be brought}into color and

gray

balancemuchmore

quicklywith _{corresponding}reductionsin

make-ready

time and paperwaste. Anotheradvantage attributedto GCR

printing

isthe

lessening

ofink

trapping

problems.With the reduction oftheamount of ink

being

putdown in

trap

the GCRmethod isutilized, sincethe most inklayersprinted would be twocol

ors plus

black,

as opposed to a fourcolorlay-down.8

There are

many

otherbenefits attributed toGCRthatare_mostly the direct result ofthe decreaseintheamount of ink

being

put onpaper. These

were presented inMichael Bruno's 1985article in American Printerentitled "Achromatics :Four Color

Printing

ThatIsn't," 9 and include :

sharper_{printing due}toall detail

being

in the

black;

reduced metameric variationsunderdifferentlight _sources; lessink consumption;

reduced

drying

problems less energy needed for ink

drying;

higher_printing speed;

the_abilitytouse lighterweight papers (thereduced inkfilm thickness should producemajorbenefits to newspaper

printing);

less dotgain and higher print_contrast;

reduced _spraypowder requirements insheetfed printing; better inkreceptivity.

Withthe obviousbenefits GCR separations afford the printer, it is

13

ColorMeasurement

When anexperimentisperformed, the results mustbe measured ina

quantitative manner. Inthe

printing

industry,

color informationhas

traditionally

beenobtained

using densitometric

measurements. Densitometers

cannot,

however,

perceive colorinthe same

way

as thehuman eye does dueto the spectral sensitivities ofthe

filters

they

use.10

"Densitometer readingswith

the conventional

filters

are therefore unsuitable foraccurate specificationof

printing ink colors unlessit is certainthat the same pigmentsare alwaysused. Eventhen, errors

may

result fromthefact thatdensitometers differ inspectral

sensitivity"11

Color _matching informationcanbemore _accuratelymeasured_using

devicesthatmeasure thespectral characteristics ofthe objects

being

compared.

Thisinformationcanbeused in its rawform (spectral reflectancecurves) or can

betransformed intounits thatrelatecolordifference ina moreintuitivemanner

(colorimetry).

Colorimetry

attempts to includeallofthe

factors

thataffectthe _waya

reproduction appears to an observer

including

spectral informationfrom the

printed sheet, theilluminantused to observe _{the object,}and a standard observer which approximates the humanvisual response system. The color space used

in this experimentis the CIELAB color measurement system. Information

concerning the theoreticalbackground ofthis system iswell

documented

elsewhere, soI willnot gointo greatdetail here. Abriefexplanation ofhow the

system was used inrelationto this experiment will suffice.

In ordertouse

CIELAB,

samples are measured

using

a colorimeteror

[image:23.569.72.499.82.539.2]

BLACK

15

information aboutthe sample. This information

may

thenbetransformed into

colorimetric values. Acolorimeter isa device that isdesigned to "see" colorthe

same as thehuman eye. The outputfrom a colorimeteris inthe units of a color spacewhich

may be

selected

according

to the needsoftheoperator. In

CIELAB,

the outputis in the formof

L*,

_a*, and b*coordinates.13 TheL*a*b*

values represent coordinates in the chosencolor space. L* providesinformation on an object'slightness and runs from zero

(black)

toone hundred (white). Informationabout anobject's hueand chroma are carried in itsa*

and b*values. The a*

value represents the object's redness orgreenness,

while theb* value representstheobject's yellowness orblueness (seeFig. 10).14 Anegative a*

valueindicates a greenhue and a positivea*

valueindicates a red hue.

Likewise,

a negativeb*valueindicates blueand a positive b*value

indicates ayellowhue.

Chroma information isalso carried inthe a*

and b*values. Thecloser

these valuesareto zero, the closerthe objectisto neutral. A

truly

neutral object

willalways havea*

and b*values of zero. For _example, ifan objectis measured

and found to havecoordinates ofL* =

80,

a*

=_-50, and b*= _-50,it is known that

the objectis

bright,

saturated,and has a cyanhue. A hypothetical

gray

object

mighthave CIELAB coordinates ofL* =

30,

a*

=

0,

andb* = 0.

Since colorcoordinates representpointsin a color space,color differences betweenobjectscanbe determined

by

calculating

the distance betweenthe

object's color coordinates withinthe defined colorspace. These colordifferences

are expressed inunits of

AE*,

which can becalculated _using the

following

AE*

=

{(LV

L*R)2 + (a*0- a*R)2

+ (b*0- b*R)2 J1^

where

L*Q

_{, a*0,} and

b*0

are the color coordinates ofthe original and

L*R

, a*R, and

b*R

arethe color coordinates of thereproduction.

The

following

example willillustratehow AE*

is calculated and howit is

useful to the printer. Atestpatch on a color proofismeasured with a

colorimeter and is found to have the

following

color coordinates: L*=

65,

a*

=

44,

b*= 27. Thejob is_thenprinted and the same testpatch onthe press sheet ismeasured. Its coordinates are: L*=

62,

a*

=

45,

b*=29.

Inorder to determine thecolordifferencebetweentheproof and the

press _{sheet, the}color coordinates are plugged into the equationabove.

AE*

= {(65 - 62)2₊ (44- 45)2 ₊ (27- 29)2}1/2 AE* ₌

{(9)

+

(1)

+ (4)}1/z

AE*

= {U}1/2

AE*

=3.74

Once theAE* valueis

known,

it is used to determinehow_effectivelythe

reproductionprocesshas matched the color of the original. AAE* value ofone is called ajust noticeabledifference

(j.n.d.)

which represents the threshold where

the humanvisual system begins toperceive color differences.

The colorimetricsystem of measurementis a powerful tool for both

17

Endnotes for Chapter Two

1 John

Yule,

"Four ColorProcesses and the BlackPrinter,"

Journal

of

theOptical

Society

of

America,

No.

30,

p.

322,

1940.

2

Gary

Field,

"Color

Variability

Associated with

Printing

GCRand

ColorSeparations,"

1986 TAGA

Proceedings,

p. 145.

3 SWOP

Handbook,

1988

Edition,

pp. 17-18.

4 J.A.S.

Viggiano,

"GCR: A PracticalApproach,"

Advance

Printing

_of

Conference

Summaries,

SPSE43rd Annual

Conference,

April

20-25, 1990,

Springfield,

Virginia:

Society

for

Imaging

Science and

Technology,

p. 204.

5

Viggiano,

p.205.

6 Dr. Abdel

Ghany

Saleh,

"Investigation into theApplicationof

AchromaticSynthesis to the

Printing

Industry," 1984 TAGA

Proceedings,

pp.152-157.

7

Saleh,

p. 157.

8

Saleh,

p. 159.

9 Michael H.

Bruno,

"Achromatics: FourColor

Printing

That Isn't,'

Colorimeter,"! 972 TAGA

Proceedings,

p. 389.

11 Pearsonand

Yule,

pp. 389-390.

12 Fred W. BillmeyerJr. andMark

Saltzman,

Principles

of

Color

Technology

,

John

Wiley

and

Sons,

New

York,

NY:

1981,

Plate IV.

'

13

Gary

Field,

ColorandIts

Reproduction,

GraphicArts Technical

Association,

1988,

p. 54.

Chapter Three

Review ofthe Literature

GCR isa topicwhich has been widelywritten about and researched. The

literaturemost often cited and most importantto thisparticular

study

is Dr.

Abdel

Ghany

Saleh's "Investigation into theApplicationofAchromatic

Synthesis to the

Printing

Industry," found inthe 1984 TAGAProceedings (p. 151).

In thisarticle, Dr. Saleh discusses the

history

ofGCR and the

theory

behind it.

Healso mentions theimportanceof _{gray balance}to theproper performanceof

GCRand investigates some of the claimed benefitsassociated withthe GCR

method of color separation.

In J.A.S. Viggiano's 1990 paper, theimportance of_{gray balance} in

determining

the

gray

componentis discussed. It is animportantconceptto

understand ifone is toperformGCR_effectively

Another important resource has been

Gary

Field'sTAGApaper "Color

Variability

Associated with

Printing

GCRand Color

Separations,"

found on

page 145 of the1986 TAGAProceedings. In this study,itwas found thatGCR

separations are_{colorimetrically}comparableto normal separations,provided

thatthe level ofGCRis low. At highlevels of

GCR,

color_variability increased.

Colorand Its Reproduction written

by Gary

Field for theGraphic ArtsTechnical

Association (GATF). It includes informationon the various color spaces

being

used inthe graphic arts

industry

and their merits.

Theseresources are the mostimportantto this study. Other

ChapterFour

Hypotheses

Gray

component replacementhas become

widely

accepted inthe graphic

arts industry. As the

technology

to performGCRseparationsbecomes more

accessiblethrough

desktop

software,it isimportantthat theeffectiveness of the

desktop

algorithmsis investigated.

Thisexperiment measured theeffectiveness ofthreecurrent algorithms

used to performGCRseparations. Theseare:Adobe

Photoshop

(desktop),

R.I.T. ResearchCorporation'sRGB-CMYKtransform

(desktop),

and Hell's first

generationGCRalgorithmincorporated inits 399ER laserscanner (high-end).

Color separations were produced_using three levelsofGCR

(0%,

50%,

and

80%)

by

eachalgorithm.

Using

eachalgorithm's non-GCRseparation as the

reference,colorvariationdue to thechange in thelevel ofGCRwasthen

determined for each algorithm. Inthis manner, color differencesproduced

by

the use ofthe differenthardware was eliminated.

Only

color variation

produced

by

_changingthe GCR levelforeach separation was examined.

Due to time

limitations,

the separationswere notprinted on _press,but

were output_using 3M's Matchprint II _proofing system. Thereare other

scanners and software packagesthatperform

GCR,

but

they

were notincluded

inthis experiment, also dueto timeand logistical considerations.

Hypotheses

HI: There isno significant color

difference,

measuredinAE*_units,between

thenon-GCRand50% GCRseparations produced _usinga Hell 399 scanner,

Adobe

Photoshop,

or RIT ResearchCorporation's RGB-CMYKtransform.

H2: There is no significant color

difference,

measured in AE*_units,between

Chapter Five

Methodology

Anexperiment was performed to investigatethe performance of some of

the GCRalgorithms used in

desktop

and high-end

scanning

systems.

Thefirst

step

in

designing

the experiment was thechoiceof a test target.

The targetneeded meet certain criteria in ordertoprovideuseful information

abouttheGCRalgorithmsused. These criteria were:

The targetmusthave a neutralscale made_up of_{cyan,magenta,} and

yellowto test theeffectofGCRonneutral areas.

Thetarget musthavea wide _variety of commonhues

including

skintonesand _memorycolors.

Thetargetmusthave both three-color and two-colorpatches

toensure that theGCR is_working whereit should be

(three-color),

and isnot_workingwhereitshould notbe (two-color).

Thetargetmustbe laid outinan_orderly and systematic

fashiontofacilitate measurement.

The test targetused forthis experimentwas the Kodak Q60AColor

Target. (Fig. 11). The Q60Awas _{specifically designed} as a colorscanner

evaluationtargetand meetsthe criteria listed above.

[image:33.569.104.471.80.370.2]

Fig. 11

-Q60AColorScanner Target

It was also engineered_using colorimetric_mapping to CIELAB aims, whichis

the color space used for evaluation ofthe output inthis experiment.1

Theexperiment was begun

by

_generating the color separationsfor each

ofthe systems

being

examined. Three separations were madewithineach

prepress _path;non-GCR,

50%

GCR,

and

80%

GCR. For the

high-end

path, the

Q60Atargetwas scanned on a Hell 399ERcolorscanner atthe desired levelof

GCRand outputto film onthe scanner's filmrecorder. Another set of

films

were then made onthe _{Hell using UCA}in the darkneutralpatchesof the target

25

Forthe

desktop

path, theQ60Awas scanned

using

an

Optronics

ColorGetter

II scanner

linked

to a Macintosh Quadra 950. Thescan was saved

as an RGBTIFF

file

which wasthenimported into the image manipulation

software for separation. The two programs investigated in the

desktop

area

wereAdobe

Photoshop

and R.I.T. ResearchCorporation's RGB-CMYK

transform.

When

performing

the separations, a problem arose as there is no

standardization

among

scanner manufacturers and software vendors for

producing

a specific GCRpercentage. Forexample, thelevel of GCRonthe

Hell scanneriscontrolled _using adialwhichhas ascale from zero to_ten, while

Photoshop

provides theusera choice offive GCRsettings: _none,

low,

_medium,

high,

and maximum. R.I.T. Research's software hadnosuchproblem as it

allowstheuser to

directly

input the desiredGCRpercentage. Since the levelof

GCRaffectsthe color characteristics ofthe _{reproduction,} comparisonsbetween

the various prepress paths are meaningless unless the GCRpercentages are

closeto

being

equal. Toachieve_this,the actuallevel ofGCR produced fora

given_setting in

Photoshop

or ontheHell had tobe measured.

Amethod of_quantifyingthe levelof GCRproduced atthe separation

stage canbe derived using thedefinitionofGCRpercentage. GCRpercentage

canbe defined as theamount of theoriginal _graycomponent replaced

by

black

divided

by

thenew

gray

component(multiplied

by

100%). These quantities can

beexpressed as dotpercentages, wherethe _graycomponentreplaced withblack

is the dotsize ofthe black

(k)

inthe GCRseparation. Thenew

gray

component

cyan, magenta,and yellowin the GCRseparation (c+m+y)/3. Given this,the

following

equation canbe obtained:

%GCR =

gray

component replaced

by

black x 100%

new

gray

component

% GCR= k _{x 100%}

(c+m+y)/3

+ k

simplifying

the denominator gives

% GCR= 3k x 100%

(c+m+y)

+ 3k

Thisequation was used todetermine the actual percentage ofGCR

being

produced

by Photoshop

and

by

theHell 399 scannerfor theirrespectiveGCR settings.

Inorder to calibrate

Photoshop,

a

gray

scale was created _usingthecolor

picker function. The level ofGCR

being

performed on each

step

ofthe

gray

scale was calculated _usingthe above equation. Itwasdetermined thatnone of

the standard GCRsettings in

Photoshop

were capable of_producingGCR levels

of 50% or80%. Tosolve _{this problem,} customGCRcurves were constructed to

produce50% GCRand 80% GCRacross the entire

gray

scale.

Asimilar process was used to calibratethe Hell 399 scanner. A

step

wedgewas mounted onthe

scanning

drumwiththe GCRfunctionturned on.

The_{scanning light}was then

manually

stepped acrossthe entire wedge while

27

Using

this method, itwas determined that a_setting of 3.5on the GCR knob

produced a close approximation of

50%

GCR. A

setting

of8 was used to

produce the desired

80%

GCR.

The

desktop

separations were thenoutput as filmpositives on theAgfa

Selectset5000 imagesetter. The films generated fromall three systems were then

outputoncommercialbase using 3M's Matchprint II_proofing system. Two

proofs weremade. One containedthe six separations performedonthe Hell

scanner and the other contained the six separations made_using the

desktop

software.

12 3

4 5 6 7 8

9101112

[image:36.569.107.476.325.618.2]

KODAK

EKTACHROME

Him

Reproduction

Fig. 12 - Areas_of the Q60Atarget measured

variation caused

by

performing

GCR.

Foreach separation_{method, the} straight scan (0%

GCR)

was used as the

control. The other separations(50% and 80%

GCR)

were compared to the

straight scan to determine theamount of colorvariation (inAE* units) caused

solely

by

the implementationof

gray

component replacement.

Eighty

patches

from theQ60Awere _measured,

including

all the simulated skintone patches, all

the three-color overprints, andbothofthe

gray

scales. (See Fig. 12 on page

27.)

All measurements weretaken _usinga

Gretag

SPM100spectrophotometer

inthe CIELAB color space. Thespectrophotometer was set_upwith the

following

measurement parameters:

Illuminant- D50

Angle - 2

Filters - ANSIT

Polarization- No

Zeroed to

-ReferenceWhite

Atotal of960 individualmeasurements were made (12targets at80

patches per target)and the datawere entered into Microsoft Excelforanalysis.

29

Endnotes for Chapter Five

TO. Maierand C.E.

Rinehart,

"DesignCriteriafoe anInput Color

In orderto test the stated

hypotheses,

namely

that there isno _{statistically}

significant difference betweenseparations performed

by

the various methods of

GCR,

thedata were _measured, entered intoa MicrosoftExcelspreadsheet, and

analyzed _usinganalysis of variance (ANOVA).

ANOVAcanbe used to test hypothesesinwhich multiple means

(u^)

of

sample populations are said to be equal, i.e.

|ij

=

\i2

= U3.1

In this study, the

means which were examined arethe averageAE* values for each ofthe

separation methodsin question. Forthis experiment, eight sample populations

were constructed throughexperimentation.

They

were analyzed in twogroups

of fourpopulations. The first

group

consisted ofthe four different separation

methods at50%

GCR,

and the second _groupconsisted ofthefourseparation

methods at80% GCR.

The first hypothesis tested,

HI,

stated/There isno significant color dif

ference,

measured inAE* units,betweenthe non-GCRand 50% GCR separations

produced _usingaHell 399 scanner, Adobe

Photoshop,

orRIT Research

Corporation's RGB-CMYK transform." This hypothesiswas tested

by

comparingthe meanAE* valuesfor eachofthe separation methods. Themeans

weretested _usinga single-factoranalysis of variance orANOVA.

31

Table 1 below gives a

summary

of themeans and variances ofthe foursepara

tionmethods at50% GCR.

Similarly,

the second

hypothesis,

H2,

states that there is no significant

difference among

the fourseparation methods at80% GCR. Table2 summarizes

these data.

Table 1

-Summary

statistics ofthe four

separation methods at50% GCR.

Separation Method

Hell UCAOff

Hell UCA On

Adobe

Photoshop

RIT Research

GCR Level AverageAE*

Variance

50% 4.79 4.60

50% 4.14 4.60

50% 1.60 3.14

50% 2.37 2.24

Table 2

-Summary

statistics of the four

separationmethods at80%GCR.

Separation Method GCR Level AverageAE*

Variance

Hell UCA Off 80% 6.81 7.78

Hell UCA On 80% 6.04 7.01

Adobe

Photoshop

80% 2.06 1.91

ANOVAtest arebeyond thescope of this

discussion,

buta briefexplanation of

how the method was used inthis experiment

may

be useful.

Theaverages and variances listed inTable 1 and Table2are the measured

result of the eightdifferent color separation methods tested. Each

grouping

of

data is a sample population or sample. The first hypothesis

(HI)

statesthat

there isno significant color differenceproduced

by

thefour separation methods

at 50% GCR.Another _wayto state thishypothesis isto_saythat the foursamples

in Table 1 all come fromthe same general population. Thesecond hypothesis

(H2)

states that thereis no significant colordifference produced

by

thefour

separation methodsat 80% GCR.

Likewise,

if H2weretrue, thefoursamples in

Table 2all comefrom the same generalpopulation.

Thegeneral population inthis experiment canbe definedas the amount

ofcolor variationproduced

by

_performingGCRon a color separation. It is

assumed that this populationfollowsa normaldistributioncentered around

somemean.

The taskis to determinewhetherthe samples wemeasured are all

subsets ofthisone generalpopulation or are members of separate and distinct

populations. WhenANOVA isperformed on multiple sample_populations,a

statistic, calledthe F-value, is generated. ThisF-valuefollows a_{mathematically}

described

distribution,

and where itfalls onthatdistributiontells the

experimenterwhether toacceptor reject thehypothesis.

Ifanalysisof variance isperformed on _anynumber ofsamplestaken

froma specificpopulation, then theF-valueproduced

by

the testwillfall

33

becomes

thatall ofthe samples in question came fromthe same population. It is

therefore

necessary

tohave a cutoff point onthedistributionat which the

experimenter can rejectthe hypothesis thatall the samples came from the same

distribution. The F-value that

determines

this cutoff is called the critical F-value

(F-critical)

and isdetermined

by

the level of confidence the experimenter wants

to achieve, thenumber of sample populations

being

compared,and thesample

size of each population.

The level of confidence canbedescribed asthe_probability of_rejecting a

truehypothesis

(a)

or of_acceptinga false hypothesis (p).

Obviously,

the

experimenterwould like to minimize the _probabilityof either ofthese types of

error.

Inthis experiment,theassumption is that the hypothesisis true, so we

areat risk of_rejectinga truehypothesis. Wewould thenwant tominimizethe

probability ofan aerror and would select a small avalue.

However,

the smaller

the a value

is,

thelargerthe F-critical valuebecomes. The possible consequence

of a large F-critical isthe acceptance ofa falsehypothesis.

A_wayto use thesmallestpossible a value and minimize the_possibility

of a

P

error isto calculatea statistic called the p-value. "Thep-valueis the

probability, given

H0

istrue, ofthe teststatistic

assuming

a value as extreme or

more so thanthevalue computed based onthe randomsample.A_relatively

small p-valuewould suggest thatif indeed

H0

istrue,the observed value ofthe

test statisticis rather unlikely. Wewould then opttoreject

H0

because thatdeci

sion would have a higher_probabilityof

being

2 In broad terms,the

p-value provides reinforcementto the decision onwhetherto acceptor reject the

rejected.3

This rule applies evenifthe F-valuereturned from theANOVAis

greaterthan theF-critical value.

The ANOVAanalysis was applied to test the twohypotheses (whichare

restated

below),

as well asto determine ifthere is a significant colordifference

between separations performed at50% GCR and 80% GCR.

HI: There isno significant color

difference,

measured inAE* _units,between

thenon-GCR and 50% GCR separations produced_usinga Hell 399 scanner,

Adobe

Photoshop,

or RIT Research Corporation's RGB-CMYKtransform.

H2:There is no significant color

difference,

measured inAE*

units,between

the non-GCR and 80% GCR separations produced_usinga Hell 399 scanner,

Adobe

Photoshop,

orRIT ResearchCorporation'sRGB-CMYK transform.

First,

let usexamine thefirst

hypothesis,

HI. Table 1 gives the average

AE*

valuesforthe foursample distributions ofthe50% GCRseparations

compared with the_{corresponding} non-GCR separations. IfHI is true, then

there isno _{statistically}significantdifference betweenthesefoursamples.

Therefore,

whenANOVA is performed,we should expectto generate anF-value

lessthanthe F-criticalvaluefor thissample size and number of samples.We

would alsoexpectto see a p-value greaterthan the avalue. Theavalue chosen

35

accepted.

Ifand

only

ifthe F-value is greater than the F-criticalvalue and thep-value is

less

than 0.05will HI berejected. Table 3 summarizes the statistics generated

fromtheANOVAperformed on these data.

Table3

-Descriptive statistics ofdata takenfrom

separations performed at50% GCR.

Sample

Hell UCA Off

Hell UCA On

Adobe

Photoshop

RIT Research Corp.

Sample Size

80

80

80

80

Sum

383.46

331.29

127.75

189.77

Average

4.79

4.14

1.60

2.37

Variance

4.60

4.60

3.14

2.24

ANOVAstatisticsforseparations performed at50% GCR.

SourceofVariation

Sumof

Squares

Degreesof

Freedom MS F-value p-value F-critical

Between Groups

Within Groups

534.18

1151.55

3

316

178.06

3.64

48.86 5.71xl0"26

2.63

The

key

valuesfromTable 3 (inbold type) are the F-valuefromthe

test,

the p-value,and theF-critical. Forthesets of separations performed at 50%

GCR,

the F-value

(48.86)

is

significantly

greater than the F-criticalvalue (2.63).

Also,

the p-valueis approachingzero.

Therefore,

I can state with

very

high

confidence that there is a significantdifferencein thecolor variationproduced

Examinationof the data generated from theseparations performed at80%GCR

yieldsthe statistics found inTable 4.

Table4

-Descriptive statistics ofdatatakenfrom

separations performed at 80% GCR.

Summary

Statisticsof separations performed at80% GCR.

Sample

Hell UCA Off

Hell UCA On

Adobe

Photoshop

RIT Research Corp.

Sample Size

80

80

80

80

Sum

544.42

483.36

164.94

195.09

Average

6.81

6.04

2.06

2.44

ANOVA statistics for separations performed at80% GCR.

SourceofVariation

Between Groups

Within Groups

Sumof

Squares

1422.38

1555.92

Degreesof

Freedom

3

316

MS

474.13

4.92

F-value

96.29

Variance

7.78

7.01

1.91

2.98

p-value

2.75X10"44

F-critical

2.63

Once again, the

hypothesis, H2,

that there is no differencebetweensepa

rations performed at80% GCRisrejected. The F-valueof96.29is significantly

higher than the F-critical

(2.63)

and the p-valueis

essentially

zero. I canreject

this hypothesis with little chance of

rejecting

a true hypothesis.

The rejectionofthe twohypothesis provesthat thereisa significant color

37

to either accept or reject a hypothesis.

In_{this case,} ithas beenshownthat the separation methods are

different

at the

twolevels of

GCR,

but further analysis of the data was

necessary

to determine

the extent ofthesedifferences.

GCR Level

Previous studieshave indicated thatcolor variationincreases as the level

ofGCRincreases. Since this

study

includes data forGCRperformed at 50%

and

80%,

it seems appropriateto examinethis phenomenon.

Table 5

-Summary

statisticsoftheseparation

methods at50% and 80% GCR.

Separation

Method GCR

Sample

Size

Average

AE*

Variance F-value p-value F-critical

HellUCAOn 50% 80 4.79 4.60 26.14 9.06xl0"7

3.90

Hell UCAOn 80% 80 6.81 7.78

Separation

Method GCR

Sample

Size

Average

AE* _Variance

F-value p-value F-critical

Hell UCA Off 50% 80 4.14 4.14 24.90 1.58xl0"6

3.90

Hell UCA Off 80% 80 6.04 7.01

Separation

Method GCR

Sample

Size

Average

AE*

Variance F-value p-value F-critical

Photoshop

Photoshop

50%

80%

80

80

1.60

2.06

3.14

1.91

3.43 0.07 3.90

Separation

Method GCR

Sample

Size

Average

AE*

Variance F-value p-value F-critical

RIT Research 50% 80 2.37 2.24 0.07 0.80 3.90

[image:46.569.75.502.390.624.2]

Since each separation method was performed at twolevelsofGCRunder

otherwiseidentical _{circumstances,}ANOVA analysis candetermine whetherthe

amount of GCR

significantly

affects the amount of color variation. Table 5on

the previous page summarizesthe colordifference data as well as the result of

theANOVAtestperformed on each separation method.

These results show thatindeed

by

traditional separation_{methods, there}

is a significantincrease intheamount of color variation as thelevel ofGCR is

increased. Bothsets of separations doneontheHell scannerhad anincrease of

over 2AE*units as theGCRwas increased to 80% from 50%.

Conversely,

neitherof the

desktop

separation methods showed a

significant increase incolordifference as the levelof GCRwasincreased. The

amount ofincrease inAE* waslessthan 0.5 inbothmethods. This differenceis

negligible invisual as well as statisticalterms.

Desktop

AlgorithmsvsHell Scanner Software

Another_way tolookatthe data fromthisexperimentis tobreakthe

separationmethods into two groups. Theseparations performed onthe Hell

scannerand theseparationsperformed onthe desktop. Table 6 summarizes

39

Table 6

-Summary

statistics

comparing

the

desktop

separation methods withthe Hell Scannerseparations.

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

HellScanner 50% 160 4.47 4.68 131.49 1.04xl0"25

3.87

Desktop

50% 160 1.98 2.82

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

HellScanner 80% 160 6.42 7.50 279.58 1.79xl0"45

3.87

Desktop

80% 160 2.25 2.47

These dataindicate that theseparations performed _using the

desktop

algorithms produced lesscolor variation atboth 50% and80% GCR than the

separations performed onthe highend scanner. The amount of variationinAE*

units is 2.49at50% GCR and4.17 at80% GCR. Thismagnitude of colordiffer

ence isvisible and would be noticed

by

an average viewer.

Comparison

Of Desktop

Systems

Thenext set of comparisons to make is the two

desktop

separation

methods. Table 7summarizes thesedata. This comparisondemonstratesthat

the two

desktop

separation methods perform similarly.

Looking

atthe

separations performed at50%

GCR,

Adobe

Photoshop

performed

slightly

better

withan average AE*0.77lessthanRIT Research Corporation'sRGB-CMYK

Transform. This difference is significant_{statistically,}butnot _practicallysince a

AE*

Table 7

-Summary

statistics

comparing

the separations

performed

by

the two

desktop

methods.

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

Photoshop

50% 80 1.60 3.14 8.94 0.003 3.90

RIT Research 50% 80 2.37 2.24

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

Photoshop

80% 80 2.06 1.91 2.32 0.13 3.90

RITResearch 80% 80 2.44 2.98

At 80%

GCR,

the two separations performance wereindistinguishable

both statisticallyand visually. Ahypothesis

stating

that the two separation

methods performed

equally

would notbe rejected _usingANOVA(F-value=2.32

and

F-critical=3.90),

and the color difference represented

by

thesample averages

is 0.38

AE*;

below the 1.00AE*

visualthreshold.

The Influence_ofUnderColorAddition

At higherlevels of

GCR,

the amount ofcyan, magenta, and yellowink

removed fromthe separationscancreate a lossof

density

which cannotbe

compensated for

by

the black ink

replacing

it. This is

especially

true inthe

neutraland near neutral areasinthe shadows. Under Color

Addition,

or

UCA,

canbeutilized to minimizethis effect. Inthis study, the high-end scans were

performed withboth UCAoff and UCAon. Table 8 shows thestatistical

41

Table8

-Summary

statistics

comparing

the separations performed on the Hellscanner with and withoutGCR.

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

UCAOff 50% 80 4.79 4.60 3.70 0.056 3.90

UCA On 50% 80 4.14 4.60

Separation Sample Average

Method GCR Size AE*

Variance F-value p-value F-critical

UCAOff 80% 80 6.81 7.78 3.15 0.078 3.90

UCA On 80% 80 6.04 7.01

Inbothcases, the average color differencewas_slightly smallerfor the

separations produced _usingUCA.

However,

thesedifferences are not

statisticallysignificant and the

differences,

measured in

AE*,

are lessthan1.0.

Endnotes for Chapter Six

1. GeorgeC.

Canavos,

Applied

Probability

and Statistical

Methods,

Boston,

MA:

Little,

Brown &

Company,

,

1984,

p. 376.

Chapter Seven

Summary

and Conclusions

The twostated

hypotheses

ofthis study,

namely

that therewould be no

difference inthe amountofcolor variation produced

by

colorseparations

performed onthe

desktop

and on high-end systems at50% and 80%

GCR,

were

rejected. Therewere, in

fact,

significant differencesbetweenthe methods. In

each case examined, the

desktop

separationalgorithms produced lesscolor

variationthan the GCRsoftware incorporated inthe Hell 399ERscanner. It is

important tonote that this doesnot reflect the color

accuracy

ofthe original

scan, _only thecolor variationproduced whenGCR is incorporated inthe color

separation procedure.

These resultslead to the conclusion_{that, for}productionworkflows_using

desktop

prepress,GCRisa valuable toolwhichdoes notintroducesignificant

colorvariationto the color separation process.

Furtheranalysisofthedata confirmed thatindeedthereis anincrease in

the amountof colorvariationasthe level ofGCRis increased. These increases

are significanton the Hellscanner,butminimal withboth

desktop

separation

methods. This

finding

is consistent with earlierresearchperformed onthis

subject.

athigh

levels

ofGCR

using

the

desktop

_methods,was below the visual

threshold of mosthumans.

Itwas also found that there was no significantdifference between the

two

desktop

separation_methods,and that theutilizationofUCAhad no

appreciable affect on color variation.

Recommendations forFurther

Study

Time andbudget limitationsprevented the test targets forthis

study

tobe

runon press. Thetargets instead were output to3M'sMatchprint

II,

an

analog

proofingsystem used to mimicthe

printing

process. It is importantto realize

that

proofing

systems (except forpress_proofs) can_onlyattempt to imitatea

press. The effect of

platemaking

and actualinkonpaperis not reflected

by

this

experiment.

Thetest couldbe repeated _usinga newer digitalscanner with a more

recentGCRalgorithm. Thealgorithm used inthe Hell 399ER was an_earlyone

and hasalmost _certainlybeen

improved,

as have thealgorithms _currently

being

used incurrent

desktop

separation software.

Also,

myskills at_running theHell scanner couldbe described as atthe

novicelevel. I have little doubtthat a professional color separator _operating one

ofthe new generationof digitalscanners could producebetterresults onthe

Bibliography

Bibliography

Achromatic

Synthesis,

Hell Graphics

Systems,

1986

Billmeyer,

Fred W. Jr. and

Saltzman, Max,

Principles _{of Color}

Technology,

John

Wiley

&

Sons,

New

York, NY,

1981.

Bruno,

Michael

H.,

"AchromaticsFour-Color

Printing

That Isn't,"

American

Printer,

January

1985,

pp.40-44.

Burgstein, Michael,

"GCR

Gray

ComponentReplacement," DuPont Photosystems and Electronic Products Department.

Canavos,

George

C,

Applied

Probability

and Statistical

Methods, Little,

Brown&

Company,

Boston,

MA.,

1984,

pp. 303-404.

Field, Gary,

"Color

Variability

AssociatedWith

Printing

GCR Color

Separations,"

2986 TAGA

Proceedings,

pp. 145-157.

Field, Gary,

ColorandIts

Reproduction,

Graphic Arts Technical

Foundation, 1988,

pp. 45-80.

Fisch, R.S.,

"Studies on the Level ofUndercolorAdditionand Black Printer Levels in GCR/UCA 4 Color Lithographic Printing,"

2990 TAGA

Proceedings,

pp.

47

Fisch,

Richard

S.,

"GCR Truthand

Consequences,"

The Prepress

Bulletin,

September/

October

1988,

pp. 6-10.

"GCR,"

Dupont

Printing

Systems

Division,

1986,

pp. 1-6.

Holub, R., Pearson, C,

and

Kearsley, W.,

"The BlackPrinter," Journal

of

Imaging

Technology,

Vol.

15,

No.

4,

August

1989,

pp.149-158.

Jackson, Lonnie,

"ComparisonofColor Lightnessin Two-Color PlusBlack

Reproduction System vs. Three-Color ReproductionSystem," Master's

Thesis,

Rochester Institute of

Technology, 1987,

pp. 1-93.

Jensen, Ebert,

"Gray

ComponentReplacement:The importance of midtone

placement and

gray

balancecontrol in four-color separationsfor web

offset/heatset

printing,"

ThePrepress

Bulletin,

September/October

1985,

pp.

12-16.

Johnson, A.,

"Practical Implementationof Optimum Colour

Reproduction,"

The

Journal

of

Photographic

Science,

Vol

32, 1984,

pp. 145-148.

Johnson, Tony,

"PolychromaticColour Removal- Revolution or

Evolution,"

1985 TAGA

Proceedings,

pp. 1-15.

Jung, Eggert, Dr.,

"Programmedand

Complementary

ColorReduction," 1984 TAGA

Proceedings,

pp. 135-150.

Kueppers,

Harald,

Color

Atlas,

NewYork: Barron's Educational

Series,

Inc. ,

1982.

Maier,

TO. and

Rinehart,

C.E.,

"DesignCriteria foran Input Color Scanner

EvaluationTest Object," 1988 TAGA

Proceedings,

pp. 469-483.

Molla, R.K.,

Electronic Color

Separation, Montgomery,

West

Virginia,

R.K.

Printing

pp. 102-107.

Pearson,

M.L. and

Yule, J.A.C.,

"Conversionof a Densitometerto a

Colorimeter,"

2972 TAGA

Proceedings,

pp.389-407.

Philippsen, Brian,

The Effectson Hue

Resulting

FromBlack

Overprinting

inHalftone

Reproductions,

Master's

Thesis,

Rochester Instituteof

Technology,

1985.

Saleh,

Abdel

Ghany, Dr.,

"InvestigationintotheApplication ofAchromatic

Synthesis

to the

Printing

Industry," 1984 TAGA

Proceedings,

pp. 151-163.

Schartz,

M. and

Holub, R.,

"Measurementsof

Gray

Component Reduction in

Neutrals and Saturated Colors," 2985 TAGA

Proceedings,

pp. 16-27.

Sigg, R,

"On Second Thought LetsCall itGCR,"T&E Center

Newsletter, RIT,

Vol.

12,

No.

6,

1984,

pp.5-6.

AppendixA

Original Data

The

following

pages contain the original L*a*b*

datacollected from the

3MMatchprintproofs. The data are grouped

by

separation_parameters, which

are

clearly

labeled atthe

top

ofeach_page, and organized

by

patchIDonthe

Q60 colortarget.

UCA

Off,

0% GCR

Patch ID L*

a!

b_I

Patch ID

L!.

_a!

b_!

1 91.59 .76 -1.00 J5 36.92 -23.39 16.28

2 88.22 3.93 1.13 K5 37.40 -27.93 12.39

3 82.79 4.98 4.20 L5 39.26 -21.75 -1.95

4 75.98 7.02 2.88 M5 36.89 -2.47 -7.94

5 70.85 6.76 3.43 A8 46.70 11.93 -5.26

6 66.95 6.03 3.74 B8 46.75 23.41 2.16

7 63.55 4.27 3.15 C8 47.11 27.76 11.46

8 58.56 5.74 4.70 D8 46.84 26.77 15.11

9 55.60 2.59 5.10 E8 50.60 22.97 21.56

10 51.73 3.03 6.61 F8 56.36 15.55 27.59

11 47.97 1.94 4.80 G8 68.10 7.17 32.83

12 45.00 1.17 4.86 H8 61.00 -8.38 26.10

13 41.73 -1.30 4.11 18 56.57 -23.50 17.06

14 37.97 -.99 4.68 K8 53.25 -29.11 12.31

15 34.81 -2.57 4.83 L8 56.56 -21.08 -9.91

16 30.68 -3.62 3.87 M8 53.30 2.15 -7.01

17 25.57 -4.49 2.99 A15 91.47 .85 -2.86

18 22.41 -5.57 1.84 B15 85.36 3.55 2.90

19 19.10 -4.93 1.40 C15 74.66 6.82 5.42

20 11.29 -2.54 -1.46 D15 66.54 7.56 5.74

A4 30.44 8.29 -21.77 E15 59.27 7.29 6.74

B4 31.32 28.59 -6.43 F15 54.28 1.53 4.34

C4 32.59 34.91 -0.17 G15 47.35 2.81 7.55

D4 32.14 31.23 11.60 H15 40.62 .21 6.35

E4 35.95 32.67 20.85 J15 35.17 -1.74 6.93

F4 38.88 17.21 26.25 K15 28.36 -1.83 5.12

G4 47.86 1.51 34.54 L15 21.03 -3.22 2.77

H4 46.41 -28.91 29.81 M15 10.28 -1.88 -1.21

J4 42.23 -45.12 23.64 A19 40.99 15.31 23.17

K4 42.83 -44.56 16.58 B19 52.06 19.63 29.80

L4 45.11 -34.57 -12.99 C19 62.91 21.66 30.37

M4 36.78 -4.32 -26.70 D19 73.35 23.67 30.27

A5 32.19 3.49 -2.80 E19 41.16 18.74 17.87

B5 31.72 14.27 -0.96 F19 51.94 23.72 23.62

C5 31.36 18.86 4.59 G19 62.35 26.42 25.52

D5 31.57 16.75 8.94 H19 76.50 20.45 23.85

E5 36.61 14.10 16.36 J19 45.49 27.26 21.20

F5 38.90 6.86 20.31 K19 56.80 26.25 22.91

G5 45.95 3.74 25.87 L19 61.76 28.19 21.36

51

Separation

Method: Hell

Scanner,

UCA

Off,

50% GCR

PatchID

LI

_a!

b_I

Patch ID

LI

_al

b_I

1 92.54 0.78 -1.29 J5 30.81 -18.80 10.16

2 89.32 3.52 -0.06 K5 30.62 -21.46 7.57

3 83.74 3.93 2.58 L5 35.15 -17.94 -3.66

4 76.80 4.72 0.33 M5 33.84 -1.79 -10.31

5 70.85 3.90 0.25 A8 45.16 10.00 -8.08

6 66.21 3.35 0.48 B8 45.78 20.08 -1.62

7 62.80 1.22 -0.30 C8 45.74 23.92 8.00

8 57.83 2.81 0.47 D8 45.32 23.03 11.84

9 55.41 0.23 0.23 E8 49.24 20.15 18.07

10 50.89 0.52 2.43 F8 56.29 11.23 23.90

11 45.77 -0.77 -0.66 G8 68.40 2.96 30.02

12 42.42 -0.8 -0.23 H8 59.30 -9.78 22.86

13 39.51 -2.55 -1.10 J8 54.14 -23.89 13.57

14 36.01 -2.37 0.60 K8 49.82 -27.33 8.80

15 32.72 -3.29 -0.59 L8 55.11 -20.07 -12.57

16 28.45 -3.22 -0.28 M8 51.70 0.84 -8.88

17 23.80 -3.16 -1.73 A15 92.95 0.19 -2.10

18 18.40 -4.21 -1.65 B15 86.54 2.76 1.53

19 14.63 -2.88 -1.57 C15 75.10 5.07 2.31

20 8.98 -2.26 -2.06 D15 66.33 4.83 2.53

A4 30.25 8.60 -23.14 E15 59.14 3.76 2.30

B4 28.89 24.75 -8.65 F15 54.62 -0.93 0.57

C4 29.16 30.53 -2.83 G15 46.00 -0.14 2.62

D4 28.71 26.35 8.55 H15 40.53 -2.14 0.92

E4 32.71 27.07 16.47 J15 34.11 -3.1 2.69

F4 35.86 13.88 21.62 K15 26.04 -2.32 1.77

G4 46.31 -2.00 32.03 L15 17.18 -2.50 0.66

H4 42.58 -27.53 25.39 M15 9.01 -2.16 -1.69

J4 35.67 -38.76 18.96 A19 39.43 11.85 16.46

K4 35.63 -37.26 12.37 B19 51.76 16.56 24.11

L4 43.12 -32.54 -13.61 C19 63.41 19.00 27.07

M4 36.14 -3.40 -27.70 D19 74.32 23.79 29.89

A5 30.02 3.11 -6.55 E19 39.32 15.40 15.43

B5 29.23 10.44 -3.75 F19 51.05 21.27 21.24

C5 28.52 15.06 1.11 G19 62.07 25.20 23.96

D5 28.53 12.34 5.18 H19 77.62 20.12 23.04

E5 33.81 10.23 11.42 J19 43.88 24.00 18.47

F5 36.26 4.01 13.94 K19 55.87 24.05 20.50

G5 44.77 0.17 20.69 L19 61.15 27.38 19.69

UCA

Off,

80% GCR

Patch ID

LI

_al

b_I

Patch ID

LI

_al

b_I

1 93.68 0.62 -0.67 J5 30.75 -16.64 8.48

2 91.39 3.24 0.94 K5 30.83 -19.01 5.12

3 86.66 4.22 3.73 L5 37.34 -16.03 -6.46

4 79.41 4.53 0.69 M5 38.21 -3.56 -10.82

5 74.39 2.17 0.30 A8 48.95 9.74 -9.27

6 70.48 1.63 -0.13 B8 48.78 18.65 -2.85

7 68.70 -0.08 -0.68 C8 48.36 22.76 5.85

8 62.74 0.92 0.40 D8 47.77 21.54 9.81

9 60.25 -1.53 -0.86 E8 52.43 19.29 15.71

10 56.19 -1.51 1.08 F8 59.71 9.78 23.24

11 53.50 -2.02 -1.17 G8 71.34 1.79 28.95

12 50.58 -2.09 -1.15 H8 61.97 -9.67 22.27

13 45.75 -3.13 -1.95 J8 55.55 -22.36 12.83

14 42.08 -3.25 -0.88 K8 51.85 -26.26 8.41

15 37.31 -3.45 -1.82 L8 57.69 -19.61 -13.84

16 32.40 -3.58 -1.86 M8 56.39 0.72 -9.83

17 24.94 -3.31 -2.31 A15 93.64 0.19 -2.18

18 20.00 -3.87 -1.58 B15 89.55 2.07 0.67

19 16.36 -3.96 -1.36 C15 79.46 3.14 1.05

20 15.39 -3.62 -2.24 D15 70.71 2.29 0.34

A4 31.39 8.42 -22.27 E15 64.37 0.92 -0.02

B4 30.02 23.09 -7.96 F15 58.00 -2.37 -1.60

C4 29.88 28.45 -2.67 G15 52.91 -1.81 -0.22

D4 28.88 23.76 6.65 H15 45.75 -3.05 -1.11

E4 34.08 25.39 15.45 J15 38.93 -3.76 0.02

F4 38.28 12.54 20.44 K15 29.57 -3.13 -0.61

G4 50.01 -2.99 31.50 L15 19.30 -3.36 -0.81

H4 42.50 -23.86 24.75 M15 15.63 -3.54 -1.80

J4 34.56 -36.72 18.52 A19 43.85 8.81 13.41

K4 34.62 -34.55 9.88 B19 55.57 13.52 22.02

L4 44.05 -30.23 -14.89 C19 66.26 17.07 26.30

M4 37.45 -3.75 -26.13 D19 75.56 22.77 28.40

A5 33.40 0.87 -7.34 E19 41.86 12.87 11.98

B5 32.53 8.38 -4.57 F19 53.65 19.87 19.08

C5 31.68 11.70 -1.26 G19 64.22 24.43 22.37

D5 31.07 10.29 3.32 H19 78.62 19.67 21.86

E5 38.33 7.63 9.47 J19 44.82 22.08 15.75

F5 41.95 1.28 13.35 K19 57.42 23.43 18.83

G5 49.76 -1.61 19.02 L19 63.10 27.37 18.15

SeparationMethod: Hell

Scanner,

UCA

On,

0% GCR

53

Patch ID

LI

al

b_I

Patch ID

LI

al

b_I

1 90.48 0.80 -0.30 _{J5 36.21} -24.40 13.99

2 87.17 3.98 1.60 _{K5 36.35} -27.30 10.32

3 81.38 5.44 4.72 _{L5 38.48} -20.52 -3.24

4 74.98 7.29 3.30 _{M5 35.95} -2.59 -7.03

5 69.51 7.19 3.70 _{A8 46.59} 12.14 -8.48

6 65.38 6.45 3.77 _{B8 46.89 22.93} -0.65

7 61.86 5.37 4.15 _C8 46.82 27.36 9.57

8 56.95 6.44 4.99 _D8 46.39 26.41 12.61

9 53.96 3.03 4.58 _{E8 50.53} 22.81 19.38

10 50.04 4.05 6.69 _F8 56.43 14.74 26.09

11 46.47 2.81 4.41 _G8 68.44 4.27 31.06

12 42.97 2.33 4.94 _H8 60.71 -10.21 24.73

13 39.94 0.23 4.29 _J8 56.14 -24.89 15.45

14 36.67 0.21 5.16 _K8 52.74 -30.49 10.48

15 32.66 -1.21 4.78 L8 55.54 -20.80 -11.01

16 28.74 -2.04 3.32 _M8 52.13 3.24 -8.16

17 24.77 -3.80 1.48 A15 90.54 1.34 3.27

18 20.35 -4.49 2.26 B15 84.53 3.07 2.05

19 17.57 -3.84 1.80 C15 74.26 6.69 4.03

20 10.41 -2.07 -0.98 _{D15 65.82} 7.00 4.58

A4 31.04 9.12 -26.30 _{E15 59.26} 5.83 4.14

B4 31.47 28.89 -10.95 F15 53.40 1.26 2.64

C4 32.36 34.58 -3.33 G15 _46.89 1.69 5.05

D4 31.72 30.29 9.42 _H15 39.81 -0.08 2.54

E4 35.85 31.60 18.83 _J15 33.98 -1.28 4.61

F4 38.94 15.98 23.67 _K15 27.15 -2.32 1.81

G4 48.06 -0.95 33.32 L15 20.40 -2.32 0.89

H4 46.77 -31.25 28.07 M15 10.06 -2.07 -2.12

J4 42.32 -46.08 22.44 A19 40.91 12.68 19.60

K4 41.79 -43.73 16.72 B19 51.87 16.93 26.59

L4 44.55 -34.24 -12.84 C19 62.84 20.13 29.65