render fotorealista

8/22/2019 render fotorealista

1/52

Beteckning:________________

DepartmentofMathematics,NaturalandComputerScience

PhotorealisticRenderingwithVray

AnjaRackwitz

MarkusSterner

15June2007

Thesis,10points,Clevel

ComputerScience

CreativeProgrammingSupervisor/Examiner:SharonALazenby

Coexaminer:Juliahln


2/52

Photorealisticrenderingwithvray|2


3/52


ACKNOWLEDGEMENTS

First,wewouldliketothankMagooasthecompanyfortakingusinundertheirwingsas interns. It has been a great and very educational experience for both of us and

withoutthatexperiencenoneofthisworkwouldbeviable.Wewerealsoassignedagreatthesisontheiraccountandlendingusbothknowledgeandequipmenttorealize

thisresearchproject.

Notonly

were

we

able

to

use

their

facilities,

we

also

joined

the

CEO

of

Magoo,

Anders,

on a trip to lmhult to visit the IKEA design studio. This leads us to our other greatsupport, Bengt Larsson, who has helped us with both the structure of the work andprovided us with a huge insight in the 3D and photography business. Other helpfulpeople at IKEA that need mentioning are Sren Larsson and Maria Forsman, whohelpeduswiththemeanstocreateourkitchenscene.Lastbutdefinitelynotleast,we

alsowould liketoextendourthankstoSharonLazenby,oursupervisorandtutorforthiscourseattheUniversityofGvleforsupportingusduringthisperiod.


4/52


Figure1:Theoriginalphotography(1)


5/52


TABLEOFCONTENTSAcknowledgements....................................................................................................................................... 3

1 Introduction......................................................................................................................................... 7

1.1 Hypothesis............................................................................................................................................. 7

1.2 Anticipatedproblems............................................................................................................................ 7

1.3 Expectedresults.................................................................................................................................... 8

1.4 Limitations............................................................................................................................................. 8

1.5 Method.................................................................................................................................................. 8

1.5.1 Methodofchoice.......................................................................................................................... 8

1.5.2 Methoddescription...................................................................................................................... 9

1.6 Questionsatissue.................................................................................................................................. 9

1.7

Purpose

of

the

research

........................................................................................................................

9

1.8 Targetgroup.......................................................................................................................................... 9

1.9 Earlierresearchonthesubject.............................................................................................................. 9

1.10 Dispositionofthepaper...................................................................................................................... 10

1.11 Typographicalconventions.................................................................................................................. 10

2 Theoreticalframework....................................................................................................................... 11

2.1 ColorandLightRealisminReality..................................................................................................... 11

2.1.1 TheNatureofRealLight............................................................................................................. 11

2.1.2 Howdoescoloremitfromobjects?............................................................................................ 12

2.1.3

Colormodels

striving

for

realism

.............................................................................................

13

2.1.4 ColorprofilessRGBvs.AdobeRGB.......................................................................................... 14

2.1.5 Colortemperature...................................................................................................................... 15

2.2 RenderingVirtualRealism................................................................................................................ 16

2.2.1 TheVrayRenderer..................................................................................................................... 16

2.2.2 GlobalIllumination...................................................................................................................... 16

2.2.3 AntiAliasingFilters..................................................................................................................... 18

2.2.4 BumpMapping........................................................................................................................... 20

2.2.5 DisplacementMapping............................................................................................................... 20

2.2.6 Highdynamicrangeimaging....................................................................................................... 20

3 VisualResponse.................................................................................................................................. 22

3.1 Photographyasastartingpoint.......................................................................................................... 22

3.1.1 Realism........................................................................................................................................ 22

3.1.2 Makingthepicturealivethreemainforces............................................................................. 26

3.2 PerceptionPsychology......................................................................................................................... 30

3.2.1 RelativeBrightness..................................................................................................................... 30

3.2.2 Colorinterpretation.................................................................................................................... 30

3.2.3

Shapesand

forms........................................................................................................................

31

4 Process............................................................................................................................................... 33

4.1 StudyExcursionDescription................................................................................................................ 33


6/52


4.2 Choosingaroom.................................................................................................................................. 33

4.3 ImageAnalysis..................................................................................................................................... 34

4.3.1 Analyzingtheoriginalimage....................................................................................................... 34

4.3.2 Thingstochangeinthe3Dscene(Figure22)............................................................................. 37

4.3.3

Postproductionin

Photoshop

.....................................................................................................

40

4.4 Experiment.......................................................................................................................................... 42

Results........................................................................................................................................................ 43

4.5 Finalimage........................................................................................................................................... 43

4.5.1 Thelittlethings........................................................................................................................... 44

4.5.2 Flaws........................................................................................................................................... 44

5 Discussion.......................................................................................................................................... 45

5.1 UsefulmeanswhencreatingphotorealisticCGI.................................................................................. 45

5.2 Pro&Contraphotography.................................................................................................................. 45

5.3

Pro&

Contra

3D

...................................................................................................................................

46

5.4 Whatcouldhavebeenmadebetter,andhow?.................................................................................. 47

5.4.1 Theimage................................................................................................................................... 47

5.5 Furtherresearchonthetopic.............................................................................................................. 47

6 Conclusion.......................................................................................................................................... 48

7 References......................................................................................................................................... 49

7.1 Persons................................................................................................................................................ 50

7.2 Websites.............................................................................................................................................. 50

7.3 Reports................................................................................................................................................ 50

7.4 Books................................................................................................................................................... 51

7.5 Tableoffigures.................................................................................................................................... 52


7/52


1 INTRODUCTION

Photorealism isonethe largestandmostcommonlyusedareas inthe fieldof3D. In

the beginning of the 21century, the Swedish furniture company IKEA started todigitallystorealloftheirproductsinalargedatabase.Inthebeginningofthisproject

alltheworkwasdonebyphotographers.However,IKEArealizedthepotentialandtheopportunitiesofthisnewgrowingmarketsimplynamed3D.Theystarted lookingfor

exceptional artists in this specific branch when they realized that their own field ofartists would not cut it. They found the company Magoo Studios this obvious step

leads

immediately

to

the

research

question.

How

does

IKEA

want

their

pictures

createdbyMagooStudiosandwhatmakestheIKEAproductstolookrealisticina3D

world? The picture will not only have to appear realistically, but also give theimpression that it was taken by a real photographer with IKEAs entire theoreticalframeworkasbackground.

Today, the pictures created with various 3D tools are not similar enough to picturestaken by a real photographer in a studio instead of in the 3D room of computer

software.Forthisprimaryreason,thisresearchisneededandveryvaluableforalotofindividualswiththesameinterest.

In this research, the main objective is to reveal the secrets of professionalphotographersandimplementtheirknowledgeoftheirtradeinto3Dsoftware.Afield

tripto

IKEAs

main

photo

studio

will

be

conducted

to

ascertain

this

information.

Most

ofthetestingwillbeaccomplishedatMagooStudiosofficetofindouthowtoreachthisgoalofaphotorealisticpictureanddevelopingaprocedurethatcanberepeatedtimeaftertimewithdifferentmodels,materialsandlights.

1.1 Hypothesis

Realism today is within the 3D creation, where not being able to tell the differencebetween two images, one made by a camera and one rendered by a 3D program.Whenyouasksomeone inthe3Dbusinesswhattheythinkwillmakeapicturemore

realistic,the

answer

is

almost

every

time:

irregularity,

dirt,

grain

or

imperfect

makes

perfect, thisconclusion isthesameoneas theonebeingbrought tothe test inthis

paper.

However,hypotheticallywebelieveitistrue.

1.2 Anticipated problems

Photography and photorealism can in many ways be a very subjective issue. This isoneproblemthatcanbecomequitelargeandtotallyunmanageable.However,inmost

cases,thisproblemwillnotoccurwhenitcomestoarealisticfeelinginanimage,but

inother

cases

it

is

not

realism

when

an

image

is

supposed

to

mediate

afeeling,

where

different viewers may have different interpretations. Otherproblems that are in thephotorealism area are the obvious problem with lighting contra material attributes.

Different materials have different attributes that work proficient with other lights,


8/52


which leads to different lighting setups for different materials, even though it is thesamemodel.Forexample,oneverydullmetallicmaterialiscreatedinthefirstsceneA.Toachievethewantedeffectonthematerial,aFresnelvaluehasbeenintroducedwhich demands a lot more intensity on the lights. Fresnel provides a more realisticfallofftothereflections,andalsofacingtheobserverspointofview.Therefore,ifweusethesamelightsinsceneB,whereallmaterialsaremuchbrighteranddonothave

asin

this

case

aFresnel

value,

everything

will

become

too

bright.

Accordingly,

every

scenedemandsdifferentlightingattentioneventhoughithasthesamegeometry.Ofcourse,evendifferentgeometrywillrequireadifferentlightsetup.

Typically,theobservercandefinewhatdoesnotappearrealistic.Thismeansthatthehumaneyeisfamiliarwitharealisticsurrounding,anddetectsevensmalldivergences

veryeasy.Thatiswhyasyntheticphotorealisticimagehastobeveryclosetorealityinsomeareaswheretheeyeisverysensitive.

1.3 Expected results

Becauseof

the

fact

that

this

paper

has

brought

up

an

enormous

subject,

no

absolute

revealingtruthwillbedeclared,althoughhopefullysomepiecestothevastpuzzlethat

is called photorealism will be put in the right places. However, this research willsummarize certain knowledge, maybe not of a global scale, nevertheless furtherinformationtohelp3Dprofessionalswiththeireverydaywork.

1.4 Limitations

SincethisresearchwillbeconductedintheserviceofMagooStudios,itwillonlyfocuson the software that Magoo has at its disposal. The software consists of the 3D

programAutodesk

3D

Studio

Max,

which

includes

V

ray

renderer.

No

comparison

of

theVrayrenderertootherrendererssuchasmentalraywillbecarriedout,becauseitisnotintheareaoftheresearchwherephotorealismwillberesearched.However,this

paperwilldescribevariousproblemsandlimitationsofthevrayrendererforrealisticimagery.Forthatreason,thisresearchispurelyartistically,nomathematicalissuesor

similarity will be researched within this thesis, as well as there will be no animationsequenceandtroubleshootingwithavoidingflickerwithanyofthetoolsthatwewill

use.Therefore,thisresearchprojectisaimedtowardMagoo,IKEA,andtheirguidelineframework,whereonlystillimageswillbeprocessedandresearched.

1.5 Method

When our research is to reach a result that is to be interpreted by the eyes, it isdoubtfulthatonemightnotcallitartistic.Justbecauseofthatfactourresearch,asalot of other research in the computer science area, will be done in the manner of aconstructiveresearch,wheretheconclusionsarebasedonempiricalfacts.

1.5.1 Method of choice

SincethisprojectisaimedwiththelineofworkthatMagooStudiosareconducting,sowillthispaper,althoughatsomewhatwiderspan.Themainresearchwillbeascenetaken by a photograph and then that photo (Figure 1on page 4) will be mimed as

similaraspossiblewiththehelpofthe3Dsoftwareatourdisposal.Whatmightbea

problem

with

this

method

is

if

the

photograph

taken

by

the

camera

is

not

exactly

like

the way IKEA wants it, the research conducted may be heading in a very wrongdirection.


9/52


1.5.2 Method description

The first step would be the recreation of the photographed scene with models andtextures, placing all the objects in the right places and so forth.The big challenge isthentolightthesceneinthesamewaythephotographerdid,withthegoaltoachievetheexactsamepictureattheendoranevenbetterresult.

1.6 Questions at issue

Howshouldacomputergeneratedimagebecreatedtolookphotorealistic?

Whattoolsareappropriatetousetoreachthewantedresult?

Whatarethelimitations,problems,prosandconsforbothphotographyand3D?

1.7 Purpose of the research

Intodayssociety,3Dgraphicsisarapidlygrowingmarketthatmostlyisestablishedin

television commercials, computer, console games, and special effects in movies. Araisingdemandforrealistic3Dfootagetherebyisadvancing.Thisleadsdirectlytoour

research, which is mainly created in the service of Magoo Studios, however also forindividuals who strive for a photorealistic result in the 3D graphics area. Due to thescaleofthisresearchandthehugetopicofphotorealism,wehavenarroweditdownaccordinglytothe limitationcaptions.Wewill focusonthecreationofphotorealisticfurnitureandhomescenes,whiletryingtomakethemasrealisticaspossiblewiththetimeandtoolswehaveatourdisposal.Withthistaskathandeverysuccessivestepinthe right direction will be documented. In short, we will make a home environment

and/or single furniture look attractive (a must have feeling), as well as appearing

realistic.

1.8 Target group

Basically, this paper is written as a foundation stone for people working withphotorealisticrendering.Itsummarizesideasandknowledgeaboutphotography,lightand color theory, and 3D, especially the rendering process. It can be useful foreveryonewhoisinterestedinoneofthesetopics.

1.9 Earlier research on the subject

Intodayscreationofactionandsciencefictionmoviesortelevisionseries,there isaveryexcessiveamountofcomputergenerated imagesusedforthecreationofeitherthings that are impossible to make with raw materials or scenes that would be tooexpensivetocreatewithouttheaidofcomputergeneratedimagery(CGI).Sinceevery

new effect or creation is different from the other, each one of these new effectsrequiressomeformofresearchtogetthedesiredresult.However,everytoolthat is

usedontheotherhandhasbeendevelopedbysomesortofresearcher;thereforeonemightaskthem,whatmakessomethingaresearch?Forexample,thecreationofthe

motion picture Lord of the rings directed and produced by Peter Jackson, hugearmies were moving across middle earth. How could such a scene be made without

CGI?To

hire

10,000

extras

and

coordinate

them,

would

be

an

almost

impossible

task

andthecostwouldbetremendous. Instead,astudioof3D individualswashiredand

accomplishedthattaskmostadmirably.


10/52


Thegoalofthisresearchistomimicaphotographtakenbyarealcamera,whichonecannotruleoutthefactofallpreviousinquiriesanddiscoveriesproducedintheworldofphotography.All3Dhasphotographyinfluences.Thereforetoexcludethehistoryofphotography research would be inevitable. To dig deeper into the category ofphotographyphysicsisalwaystobefound.Becauseofthis,alotofphysicaltermsareusedintodays3Dgraphics.

The issuethathasbeenbroughtforth inthisdiscussion isthefactthatallofthishasbeenalreadycreatedonvariouslevels,butitisalsouptooneanotherwhatyouwouldcall research. Though it might be hard to believe that anyone previous has made aresearch onhow IKEAwants theirCGIsappearance. The art director at IKEA mighthave someclear ideasorhave made somedrafts of them.However, theymight not

have lookeddeeper intothesubjectofwhy theywantaparticular look, inaspecificwayorhaveacharacteristicfeel.

1.10 Disposi tion of the paper

Photorealism

in

images

can

be

seen

as

the

art

to

find

a

balance

and

harmony

betweenscientificstructuresandpureartisticwork,saysBengtLarsson.Tosuccess

withthiscombinationwesometimescallafeelingforimages,directlytranslatedfromSwedishsbildknsla(2).That iswhythisresearch isdividedwiththescientificpartabout the theoretical background, whereas the artistic side of our work will bediscussedthereafterinvisualresponse.Thepracticalwork,describedinpart3,consistsofcloningaphotoofoneofIKEAskitchenin3D(Figure1).

1.11 Typographical conventions

Manyof

the

translated

technical

words

are

taken

from

awide

ranging

dictionary

on

theInternet.(3)

Thesoftware3DStudioMaxabbreviatesto3dsMax.

Threedimensionalabbreviatesto3D.

HighdynamicrangeimagingwillbeshortedtoHDRI.

ComputergeneratedimagerytoCGI


11/52


2 THEORETICAL FRAMEWORK

The process of creating synthetic images that are indistinguishable from realphotographs is called realistic image synthesis (4). To achieve photorealism with acomputerisachallengingtaskandrequiresunderstandingofthefundamentalphysicsandpsychophysicsof light.Howdoes light interactwithmaterialsandsurfaces inthe

realworld?Whathappenswhenlightraysenterthehumaneye?

In the 3Dworld there are many mathematical algorithms that simulate thisprocess.Light is bouncing around from a light source through the scene into the camera.

Depending

on

the

objects

materials

the

light

creates

reflections

and

refraction

on

surfacesor isscattering inamedium.Thereareseveral importanttermsandtoolsof

photorealismsuchasvray,globalillumination,raytracing,antialiasing,andsoforth.Sincetheyarekeystonesinthisresearchathoroughexaminationofthesesubjectsisinorder.

2.1 Color and Light Realism in Reality

2.1.1 The Nature of Real Light

TheGreekshavefoundedthebaseoftodayscurrenttheoriesof light.Theybelievedthat

light

was

emanating

from

the

eye

and

touching

the

objects

that

we

see.

Even

first

theoriesaboutreflectionandrefractionweredescribedby theGreekmathematicianEuclid, the astronomer Cleomedes and Ptolemy, mathematician, geographer,

astronomer, and astrologer. New breakthroughs in understanding light were merelycompleted at great distances until the 17

th century with Christian Huygens (1629

1695), a Dutch mathematician, astronomer and physicist and Isaac Newton (16421727),theregardedEnglishmathematician,physicist,astronomer,naturalphilosopherand alchemist. Many breakthroughs in mechanic, gravitation, and optic were madeespecially by Newton. However, since their findings, light was interpreted as a wavemotionandcouldbeseparatedoutintoitscolorspectrumwithaprism.Newtonalsopublishedatheorythatlightparticlesareemittedfromlightsourcesandmoveinstraitlinesuntilthey impingeonasurface.ThomasYoung(17731829)andAugustFresnel

(17881827)

studied

the

effects

of

polarization

and

diffraction.

That

gained

credence

to

thewave theoryrefinedby JamesMaxwell (18311879),whodefined the lightasanelectromagneticwave.Intheearly20

thcentury,AlbertEinsteinmodernizedthewhole

physicalworld,amongothersbyintroducingtheuseofphotonstodescribethephotoelectric effect; the process whereby electrons are liberated from a metallic material

which is rayed with electromagnetic radiation such as xray, ultraviolet light ordaylight.Thephotonistheelementaryparticleinlightexplainingtheelectromagnetic

effects.


12/52


Today,thephysicsoflightisdefinedasacombinationofseveraldifferentmodelssuchas ray optics, wave optics, electromagnetic optics, and photon optics. In computergraphics,ray optics is the fundamentaland almostsolely model for lightcalculation.(4)(5)

Figure2:Theelectromagnetic spectrum,whichencompasses thevisible regionof lightfor the

humaneye

2.1.2 How does color emit from objects?

The process of seeing is much more complex than taking a picture with a photocamera.Thepictureon the retina in theeye is notonly thereflection of thereality.Thispictureneedstoconverttoelectricalimpulsesandtobeinterpretedbythebrain.

Light falls onto an objects surface. The light is absorbed, refracted or reflected indifferentwavelengthsfromthesurfacedependingonwhatkindofmaterialtheobjectconsistsof.Thenthiscertainwave lengthenterstheeyeandreaches theretina.Thehumaneyehasatrichromaticcolorsense,whichmeansthreedifferenttypesofcone

cellsfor

red,

green

and

blue

light

that

that

turns

the

incoming

information

to

neuronal

signals.Ifalltheformsofconesarestimulatedequally,thecoloriswhite.Blackisseen,

if none of them responds. This combination offers to apprehend the lights wavelengthsfrom400nmto780nmofthecolorspectrum(Figure2,Figure3).Forhumanbeings, this color spectrum is continuing and means that they can apprehend every

color and light, but ultraviolet and infrared light. Some animals have this ability, forexample bees. They are sensitive for ultraviolet light for everything but the redfrequencies.Thebeesseeredasgreytones.

Therefore, color can be defined as both subjective (what we see on an objects

surface),andevenobjectiveas thecertainwave lengththathasbeenreflected fromtheobjectssurface.(6)(7)

Figure3:Thisimageshowsthevisiblespectrumofwhitelight


13/52


2.1.3 Color models striving for realism

The observers color impression is subjective. Everyone apprehends different colors.Thepresentdaylivingischaracterizedbyallthinkableformsofimagesinmedia.Thereare both printed media like papers, books, magazines, posters, catalogues, etc andnonprintedmediasuchastheweb,film,computergraphicsinvideogames,computergames,presentations,advertisingandmuchmore.Toproducephotorealistic images,

the reproduction of the original will, appear exactly the same in an ideal case. Theprocessforthemajorityofeveryreproductionforscreenorprintisdigitaltoday.This

requires technically that all equipment display the colors in the exact same way,practicallythisisnotpossible.

Thedifferencebetweenprintedornonprintedpresentsthemainproblem.Printscanonlyabsorborreflectacertainwavelength(subtractivecolormixwithcyan,magenta,yellowandblack),whilescreensofanykindareworkingwithanadaptivemixoflight

(phosphors).Theimaginationintheidealcaseisforexampleamonitorcandisplaytheequalcopyofapicturewith itscolorand intensity.Assoonas the light in theroom

changestodark,theoriginalpictureloosescontrastandintensity,whiletheimageonthescreenappearstobebrighter.Theoppositehappenswhenlightmeetsthescreen

andthe

contrast

nearly

disappears.

The

human

power

of

seeing

is

very

adaptive

and

adjustssuchvarianceincontrasts,color,intensityetc.

Even equipment working with the same technology can heavily distinguish themselves. These differences are often based on the equipment construction andfunctionality. A monitor uses phosphors to produce colors. A scanner or digital

camerashassensitive lightsensorstocapturedata.Printerscolorsaregeneratedbydifferent pigments of CMYK color model. For exampleworking with colored pictures

withascanner, a monitor, andaprinterwithoutusing the sameColor ManagementSystem(CMS) involvesgreatdiscrepanciesofcolorvalidity.Therefore,everydeviceisdelivered with its own machine profile, which has to be compatible with others,following the International Color Consortium (ICC) standard. Every machine workingwiththe imageshouldusethesamecolorprofileandeventhesamecolormodel,as

longas

possible.

With

every

conversion

from

one

gamut

to

another

(for

example

RGB

toCMYK)somecolorinformationwillbelost,becauseoftheirdifferentcoloramount.(6)

PrinterscanonlyworkwiththeCMYKcolormodel,whichissmallerthantheRGBcolormodel.Thatcreatessomeproblemsforsomecolorsarenotprintable.Theyhavetobeconverted to an equivalent CMYK color mix. It appears always some divergence,especially luminous colors. Very bright white tones or very dark black tones arecomplicatedtoprint.Thisisaboutupto34%colorinwhiteormore95%ofblack.Theprinters cannot display this minimal amount of color and a sharp edge appears.Therefore, IKEA sets a limit to 247 for white (maximum 255) and 10 for black whileworkingintheRGBmode.Thislimithastobeobservedespeciallyforborderregionstoprevent the object from floating out into the whitebackground.Smaller regions like

highlightsinreflectionsaregrantedexception.


14/52


2.1.4 Color profiles sRGB vs. Adobe RGB

The human eye can apprehend a color spectrum as represented in the CIE xy

chromaticitydiagram(Figure4).Thisspectrumismuchlargerthananycolorprofile.AcolorwiththreeRGBvalueshasalsotobeassociatedwithacolorprofilelikesRGBor

Adobe RGB as two of the most common representatives. sRGB was created in

cooperationby

Hewlett

Packard

and

Microsoft

Cooperation

for

certain

use

on

LCD

and

CRTmonitors,digitalcameras,scannersandprinters. Ifproperlycalibrated, theycannearlyproduceallcolorsinthesRGBscolorgamut(colorspace).ThesRGBsgamutisthe smallest one and therefore limited for more demanding outputs such asprofessionalphotoprints.WorkingwithAdobeRGBoffersawidergamutespeciallyforgreen tones as shown below (Figure 4). However, due to the fact that those colorscannotbedisplayedonusualscreens,thisprofiledemandsbetterandmoreexpansiveequipment.AssigningthesRGBprofiletoan imagewiththeAdobeRGBprofilecould

result in loosing color information. (8) Rendered images, delivered by 3ds Max, areuntagged. Their color spectrum closely resembles the sRGB gamut with its color

structure.Thereforewecanhandletheminafairlypredictableway.(2)

Figure4:ThisimagesshowsthesRGBgamut(thesmall)andtheAdobeRGBgamut,bothplaced

in the CIE 1931 color space chromaticitydiagram. The outer curved boundary represents the

spectralcolors,withthelightswavelengthmeasuredinnanometers.


15/52


2.1.5 Color temperature

Principallyeveryobjectreflectsacertainwavelength,butnottheidealizedblackbodywhich absorbs any kind of light. A light sources color temperature meters thetemperature that a black body has while lighting it with different lights. The colortemperatureismeasuredinKelvinasinsomeexamplesinthetablebelow(Table1).

CommonColorTemperatures

Source K

Candleflame 1900

Sunlight:sunsetorsunrise 2000

100watthouseholdbulb 2865

Tungsten lamp (500W 1k)

3200

Fluorescentlights 3200

7500

Tungstenlamp(2k 10k) 32753400

Sunlight: early morning,lateafternoon

4300

Sunlight:noon 5000

Daylight 5600

Overcastsky 60007000

Summer sunlight plus sky

blue

6500

Skylight 1200020000

Table1:Realworld color temperatures, startingwith the lower temperaturesgoingfrom red,

throughwhitelighttoblue.(9)


16/52


Thelightisredifthecolortemperatureisaround2400K,yellowat4800Kandbluishat9300K.Daylightisdefinedaround6500K(Figure5).(10)

Figure5:ThecolortemperatureplacedintheCIE1931colorspacechromaticitydiagram.

Evenwhenknowingallaboutcolor,colormodelsandprofiles,colortemperature,etc,

still leaves hundreds of questions unanswered. Bengt Larsson expressed himself thisway:Formecolorseemstobeamixofvoodooandmarshyground.(2)

2.2 Rendering Virtual Realism

2.2.1 The V-ray Renderer

Vray

is

arendering

plug

in

for

the

3D

software

3D

studio

Max

which

supports

almost

every feature in 3Dstudio max. The plugin is made by a Bulgariancompany namedChaos group.Vray is an extremely powerful tool when it comes to rendering. Vrayhas large arsenal of tools to its disposal, for example ray traced reflections, indirectillumination,caustics,antialiasing,etc.Therewillbeashortdescriptionlaterofsome

of theexcellent tools that isprovidedwith thevrayplugin,whichalso involvesourresearch.(11)(12)

2.2.2 Global Illumination

Global illumination or GI is yet another tool for photo realistic images. Global

illumination is a set of algorithms that can be used in 3D, for example Ray tracing,beam tracing,ambientocclusionorphotonmapping.Allof theseGlobal illumination

methodsuse

algorithms

named

diffuse

inter

reflection

and

specular

reflection.

Diffuse

interreflectionisdescribedbylighthittinganunevensurface,bouncesoffitagainandhits other surfaces, thereby illuminating them as well. While specular reflection is a


17/52


reflectioncomingfromaperfectevensurfacesuchasamirrororalakeonacleardayandthereforereflectstheraysinastraightangleaccordingtothelawofreflection.Thislawstatesthatanyincominglightthatbouncesoffasurfaceandisreflectedwillhavethesameangelastheincominglight.(13)

The main goal of global illumination is to enable a correct calculation of the lights

intensity

at

any

point

in

the

model.

Without

using

global

illumination

the

image

often

appearsflatandsyntheticandloosesitrealistictouch,forexamplecolorbleeding.Thecoloringofthesurfacefromthereflectionsofthespherescanbesimulatedwithglobalillumination,otherwisewithradiosityorcoloredlight(Figure6)(14).

Figure6:Arenderwithout(l)andwith(r)globalilluminationon.

Themajorityoftheglobalilluminationalgorithmsthathavebeendevelopedarebased

ontwomajortechniques:Raytracing(pointsampling)andradiosity(finiteelements).Bothofthemhavetheirstrengthsandweaknesses.Hybridtechniquesarecombining

thebest

of

both.

(4)

2.2.2.1 Ray tracing

Ray tracing is a method used mostly when creating photo realistic images. However

because of its precision and excellent quality, it also is a very demanding renderingmethod.Whattheraytracerdoesissimilartothename,whichismisleading.First,theraytracerssoftwaredeterminatesthepositionofthecamerainthescene.Secondly,itcalculatestheanglewherethecameraispointedanditsfieldofview.Afterthatphase,the ray tracing algorithm sendsouta ray for each pixel in the viewport out into thescene.Thenthesoftwarecalculatesthecolorofeachpixel,dependingontheraysentfromthecameraandbouncedonthetargetedsurfacethatithitslater.Raytracingis

thoughttobethemostaccuratemeasureofrenderingpossible.However, italsohasflaws, such as depth of field. It cannot be rendered, because with a ray tracing

algorithmeverythinggetsperfectlysharpandeven.(15)(16)


18/52


2.2.2.2 Radiosity

Radiositytechniquesweredevelopedand introducedbyateamofresearchersattheCornellUniversityat1984;someyearsbeforetheraytracingmethodswasestablishedasanalternativemodel.Thisalternativefiniteelementmethodisaglobalilluminationalgorithm handling the interaction of light on purely diffuse surfaces. This methodsinnovation consists in objecttoobject reflections. Earlier light reflection models did

not account for the interaction of Lambertian (diffuse) surfaces and thereforeincorrectly calculated the global illumination effects. The surfaces in the scene are

divided up into smaller patches, which are rendered in turn. For each pass of thealgorithm, the total amount of light at each patch is calculated, meaning that light

bouncingoffasurfacehittinganotherisaddedaswellinthenextpass.

Laterradiositycouldhandlemorecomplexreflectionsmodels,howevernospecularorglossysurfaces.Radiosityisquiteefficientinsimplesceneswithdiffusematerials,but

itbecomes verycostly in termsof rendering timeandstorage requirementswhen itcomestocomplexmodelsandnondiffusematerials.(4)(17)

2.2.2.3 Caustics

Caustics are called the reflection patterns that appear when light is reflected on ametallic surface or refracts through transparent objects such as glass, water or ice.These patterns depend only on the light sources position as shadows, not theobserverspointofviewsimilartoreflections.Intherenderingbelow(Figure7),therearecausticsontheflourandintheglasses.Thisimageisgeneratedwiththeaidofavraytutorial.(18)

Figure7:Especiallywhenrenderingglassandmetallicmaterialcausticscanbegenerated,asin

thisimageonthefloor.

2.2.3 Anti-Aliasing Filters

Originallyaliasingisanaliasofoneuniquesample,whenyouusethesameinformationover and over again, a pattern of these identical parts may appear if you have notsampled the original information high enough, and form a distortion in the image

edge.Thisproblemcommonlyoccurswhenanimageisgivenalowerresolution.Antialiasing is exactly what it implies, a counter measure to these aliasing effects, a

mathematical cure to the problem with undersampling or low resolution images.

Whatan

anti

aliasing

filter

actually

does

is

that

it

distorts

or

blurs

the

original

shape

edge, fooling the eye where it is more exact than it actually is. There are manydifferenttypesofAntialiasingfilters,foundintherenderingwindow.


19/52


Someof themjustblur thejaggededges,othersjitters theoriginalpixelwhileothermethodschangesthetoneofthecolortone.

Vrayoffersbothsharpeningfilteraswellassofteningfilters.Thedefaultfilter isthearea with a variablesize filter. The parameter can change from 1.0 to 20.0 where

smaller size means a sharper result, as for example 1.5 which is at default. The

Blackman

works

equal

to

the

sharpen

filter

in

Photoshop,

though

without

extra

parameters to control. The CatmullRom gives a similar sharpen effect like theBlackman, but it even slights the edgeenhancement. There are no controllableparameters in this filter either. Another sharpening filter is SharpQuadratic 9pixelreconstruction filter from Nelson Max. Especially in animation sequences it is notalways wanted to sharpen details, but provides a smoother look. The filters such as

Cubic,Quadratic, Soften orVideo can be applied. Some filters can both soften andsharpen,forinstanceCookVariablewithvaluesof1.0to2.5forsharpeningandhigher

valuestoblurtheimage.EvensoMitchellNetravaliisworking,wheretheuserisgiventovalueanisotropytoblurandringing.Bothvaluesvanvaryfrom0.0to1.0while0.33

isdefault.WhilerenderingshotsusingdepthoffielditcanberecommendedtoapplyBlend,sincethisfilterblendsbetweensharpareasandGaussiansoftenfilters.

It

often

takes

some

experimenting

time

to

achieve

the

best

result

for

a

particular

image.(9)

2.2.3.1.1 QuasiMonteCarlo

Monte Carlo is an algorithm simulated by computers and used in a wide area ofexpertise. This simulation is most useful when calculating things that have a largedegreeoffreedom,suchascellularstructuresorfloatingliquids.MonteCarlohasalsoproven to be very valuable in computer graphics, whilst calculating the globalilluminationbecauseofitsuniqueabilityofrandomlyfindingthecorrectresult,aswellas being very precise. The only downside in this context would be the highlydemandingcalculationtimes.

QuasiMonte

Carlo

on

the

other

hand

is

pretty

similar

to

the

regular

Monte

Carlo

method. Thedifference is that the Quasiversion works with numbers that are in anorderedsequence,whiletheregularoneusespseudorandomnumbers,whichmeansthat the numbers seem to be random but are not. This is why the Quasimethod isfasteringeneratingaresult.

AswritteninthehelpfilesfortheVrayrenderer:TheQuasiMonteCarlomethodfor

computingglobalilluminationisabruteforceapproach.Thisisextremelytrue.WithintheboundariesofGlobalillumination,theQuasiMonteCarlomethodrecomputesallpixels in the whole scene that are shaded and calculates upon it individually. Thismightseempeculiar;however,itisalsoextremelyaccurate.(11)

2.2.3.1.2 Photonmap

As previously mentioned, radiosity and raytracingbased methods are combined inhybridtechniques,sinceradiosity isefficientatdiffusereflectionwhereasraytracingcreatesgoodspecularreflection.Photonmappingtakesanalternativeapproachthanthehybridtechniques.Withphotonmapping,theinformationisstoredaspointsinthe

photonmap,whichisaseparateindependentdatastructure.Thelightsemitphotons

thataretracedthroughthescene,reflected,refracted,absorbedorscattered,buildingthe photon map that contains information about all photon hits. Especially in verycomplexmodels, thisprocesscalledphoton tracingsimplifies therepresentationandmakes the rendering process more efficient. This advantage is based on thefunctionality of the photon maps illumination that is decoupled from the geometryand allows handling even arbitrary surfaces. The photon map can be seen as a light

cache inbidirectional path tracing, which means sampling the image both from the

lightsources

(such

as

caustics)

as

well

as

from

the

observers

point

of

view

(such

as

mirrorreflections).Inthesecondpassthe imageisrenderedusingtheinformationinthephotonmap.(4)


20/52


2.2.3.1.3 LightCache

Lightcachingalsoknownaslightmappingisaprocedurethatisverysimilartophoton

mapping,although,itisconsideredtobeitsreplacement,becauseofitssuperiorityinmostcases.Lightcaching issimilartophotonmapping,but insteadofcalculatingthe

lightemittedfromallthelightsources,thepathistracedfromthecameraandintothe

3D

structure.

Then

the

accumulated

energy

is

stored.

Advantages

using

light

caching

instead of photon mapping are for instance a functioning use of skylights and omnilights.Lightcalculations incornersseemmorecorrectand inmostcasesthepreviewresultsseemveryquick.Althoughsimilartothephotonmap,lightleaksmayappear,aswellastheuseofbumpmappingdonotworkeverytime.(11)

2.2.4 Bump Mapping

Bumpmappingisamethodwherethecamerahasbeentrickedtointerpretasurfacedifferentfromwhatitactuallyseemstobe,withthehelpofabumpmappingtexture.

Applingaspottedtexturetoasphere, itwillappearsimilartoaballwithmanysemideep holes in it. There are two methods for Bump mapping in the field, real Bump

mappingandfake.

The

real

Bump

mapping

method

takes

each

pixel

from

the

Bump

map

texture

and

calculatesaheightmapfromit,penetratingthesurfacenormalandbinormalateachrenderedpixelofthesurface.

ThefakemethodsofBumpmapping,alsocalleddot3Bumpmapping,isoftenusedin3Dgraphicsforgames. ItworkssimilartoNormalmapping,because itusesatexture

mapwithcolorinsteadofagrayscale,deliveringalotmoreinformation,whichontheotherhandismoretimeconsuming,too.Thecolorsblue,green,andredrepresentthenormals x, y,andz in 3d space.Eachpixel in the Bumpmap texture is treated as avertex, which is placed upon the material, where the object has a whole newappearance.(19)(20)

2.2.5 Displacement MappingDisplacementmappinggivesasimilarresultasbumpmapping,exceptitdoesnotalter

thewaythecamerainterpretthenormals.Instead,italterstheactualgeometry,whichcan cast shadows, occlude other objects, basically everything a real geometry does.

The first renderer available, that could create a Displacement map, was PixarsRenderman.TheRendermanrendererusedamethodcalledmicropolygonrendering.

Every pixel in the texture for the Displacement map was treated as a polygon. Inprograms such as zbrush, this map for the Displacement can be created and then

appliedtotheobjectinthe3Dprogram.(21)

2.2.6 High dynamic range imaging

Thehumaneyeworksveryefficientwhen itcomestofastadaptiontochanging light

intensities.It

has

the

possibility

to

distinguish

millions

of

detail

in

color

and

intensity

bothonthebeach,abrightsummerday,andinadarkroomlitupwithcandlelights.Even if thedaylight is much brighter than thecandle light, theeyecanadapt to theintensity. This adaption makes it possible to see details both in a dark room andoutside the bright window. A standard camera cannot adapt to several differentexposuretimes.Thatiswhytheimagehaseitheroverwhiteareasasinthewindow,

oroverblacksinthetoodarkroom.

Today,HighDynamicRangeImagingorHDRIisafrequentlyusedlightingtechniquein3D,whichmeansgettingthemaximumofcontrastinapicture.AnHDRimagehasanextra floating point value associated with each pixel that used to define thepersistenceoflightatthatpoint.Whenitcomestodescribingthelightvaluesperpixel

precisely,

a

low

dynamic

range

image

(like

every

usual

jpg

picture)

has

a

limitation

of

250:1. The dynamic range for human eye is about 10.000:1. Some of the HDRIs canhavearangeofover100.000:1.Theonlyproblemisthatacommonscreen,monitororprinter cannot reproduce this high dynamic range because of their limitation to 256


21/52


intensity values. Therefore, you have to convert the HDRI to a LDR with a processcalled tonemapping. This includes a range of different methods to get an infinitedynamicrangetoalimited(0or1).(22)

AnHDRimagecanbetakenasseriesofseveralpictureswithchangingexposuretimes.

Ithastobeastillpicture,photographedfromthesameposition.Theshiftingintensity

in

the

pictures

from

over

white

to

over

black

demands

the

compositing

process

afterwards.

HDRI is mainly used for coherence because of its ability to mime the real world,especiallyusefulwhenworkingwithreflectiveobjects.Forinstance,anHDRIimagecan

betakenoftheenvironmentandthenbeappliedtotheobjectsscenein3D.

ItcanalsobeveryusefulforImagebased lighting,wheretheHDR image(Figure8)is

used for emitting the light or photons from a sphere around the scene. It is oneapplication where a high range of color and intensity is required to illuminate the

sceneinarealisticway.Oftenthoseimagesarealsopanoramas,tobeemployedonasphere.Sucha360angledimagescanbetakenphotographingaspecialmirrorsphere

that reflects the whole environment, but not the camera itself after retouching thepicture.Becauseofthespheresroundingtheimage,theybecomestretchedsimilartothe

fish

eye

effect.

Then

the

HDR

image

has

to

be

mapped

on

a

sphere

in

the

3D

scene.(23)

Figure8:AHighDynamicRangeimageforimagebasedlighting


22/52


3 VISUAL RESPONSE

In this chapter, visual response, the artistic side of imaging will be analyzed deeper.Whatmakesapictureappearsgoodorcomfortable,interestingandmodern?Whatofdoestheimagesharmonyconsist?Whatismeantwithafeelforimages?

3.1 Photography as a starting point

Duringour

research

period

we

obtained

the

opportunity

to

visit

IKEA

Communications

AB, inlmhult,Sweden. IKEACommunications istheheadquartersforall IKEA imagesynthesisandphotography.Mostofthephotographicaltheory,suchassettingupthe

lighting,isbasedonthephotographersexperience,andthereforedoesthefollowingstudymostlyreflectIKEAsfeelingforimages.

3.1.1 Realism

3.1.1.1 Lighting

3.1.1.1.1 Basicthreepointlighting

Lightingcan

virtually

be

seen

as

amodeling

tool

that

is

why

the

three

point

lighting

has

becomestandardinlightingthreedimensionalforms.Asthenameimplies,thelightingsetupconsistsofthreelights,withaspecificfunctionforeachoneofthem.

The key light or main light defines a scenes dominant lighting with the highestintensity of the three lights and also casts the main shadows. It generally is placed

abovethesubjecttosomedegreeandbesidethecamera.Ontheotherhand,aroomsetmayrequirethekeylightatthepositionofabrightwindow,illuminatingthescene

withdaylight.Evenplacingitbehindtheobjectcanbedesirabletoemphasizeamoredramaticfeeling,presentingonlytheobjectssilhouette.

Thefill lights main function is to light up the scene from the key lights oppositeposition. This light is especially needs to open up the key lights shadows, reducing

theirdensity.

Rendering

with

ray

tracing

and

global

illumination

already

includes

that

lightisbouncingbackonsurfaces,indirectilluminatingthesceneasthefilllightdoes,butevenmorerealistically.

More depth can be achieved with the help of a back light, which is essential forseparatingtheobjectfromthebackground,especiallywhileworking insimilartones.

This light is can also be called as hair light or shoulder light, when filming portraits,givingatouchofgloworhighlighttotheedges.

In thephotostudio, the fill light and the back lightcanalso bejustreflectorplanes,which open up shadows or reduce some bright, unwanted reflections. These planesareoftenwhiteorblack,butevensilvercanbeusefulforsecondarylighting.(9)

3.1.1.1.2

Hardandsoftlight

Different types of light intensity and power cause different expressions.Hard lightcreatesastrongcontrastwithbrighthighlightsanddark,sharpshadows.Direct light


23/52


from a lamp can often generate such hard light, which often provides suspense anddrama and intensity to the scene. An indirect lighting creates softer shadows andsofterhighlightsontheobjects.Thiscanbeachievedthroughturningthelamparoundand illuminate a big white or metallic reflector, where the incoming light from thelampbouncesoffandintothescene.Especiallyforacozyindoorfurniturescene,suchlightingwillbedesirable.Andespeciallyforsceneswithwindowlightingupthescene,

indirect

light

is

more

realistically,

since

the

outdoor

light

is

bouncing

around

a

lot

beforehittingthewindow.

3.1.1.1.3 Warmandcoldlight

Asseenin IKEAshugephotostudio,thephotographersactsimilartoartistsuntilthefinalpictureistaken.Theyaremodelingwithlighttoachievethebestcontrast,depth

andfeelingintheimage.Butnotonlyreflectorsandwhitelampsareessentialforthecreationofthe image,buteventhe lightscolortemperature.Havingacloser lookto

the white lamps with special measuring instruments, quite huge aberrations fromwhite can be detected. In some cases the color temperature is even shifting from

bluish to yellowish, for instance. While the photographer may be is struggling withtheseeffects,themathematicallyexactlightin3Dmayneedafewmoreofthoselivelylights.

As

the

photographer

can

experiment

with

several

colored,

transparent

plastic

foils,thesyntheticlightsin3Dcanchangecolor.Asdescribedbefore,thelightscolortemperature varies a lot depending on its type of light source, such asdaylight with6500 K, as white light, or a candle light with about 1900 K, as very red and warm

light.

Thewholetricktoachieveaharmonicbalance inthe image istomixcoldandwarmlights(2).Coldorvery light isoftencomingfromoutdoor,thoughawindowordoor,

for example while lamps of the brand tungsten (between 32003400K) in indoorlighting spread warmer light, creating a balance between blue and yellow, or cold

andwarminthescene. Itwillfurthermoregenerateafeelingofdepthandspatiality,astheplaywiththebacklight.

Figure9:Usingwarmandcoldlightinanimagegeneratesafeelofdepth.

3.1.1.1.4 Indirectanddirtylight

Withoutevenhavingturnedononesinglespotinalightsetup,socalledindirectlight(2) from the surrounding, the photo studio is bouncing around and lighting up

everything.Thatisalargeamountoflightincomparisonwiththevirtual,totallyblack3Dscene.This indirect lightcomesfromother lightingsetups,reflectedonwallsand

ceilings or windows, spreading daylight, and is nearly impossible to control. Though,thislightcanliftupthesceneimmensely,makingitmorealive.

Evenwiththebestandmostexpensivelightingequipment,alampcannevergenerateatotallyevenspectrumoflight.Therearealwaysminimalirregularities,whenthelightfromthelampisshiftingfromwarmtocolderlight,whichcanprovidesomeusefulandsuggestiveunbalances.Thislightcanbeseenasdirtylight.


24/52


3.1.1.1.5 Shadows

Shadowsareessentialfortheimageandtheimagesexpression.Withoutlightthereis

onlyshadow,andwithoutshadowthereisnodepth.Inarealworldscene,lightalwayscastsshadows. In3Dtheshadowcanbeturnedoff,whilethe light ison.Those light

settings can create excellent artistic effects, although they can cause an unrealistic

feeling

in

the

image,

too,

because

of

the

lack

of

shadows

where

the

observer

expects

them.

3.1.1.1.6 Reflections

Theusageofreflectionsinphotorealisticimagesisveryimportant,becauseoftheeyes

amazingabilitytonoticeirregularityandthingsthatseemoutoforder,suchasaglassthatwillnotbereflectiveatallorreflectssomethingthatistotallyoutoforder.Thisis

anexcellenttopictoincludeHDRI.BecauseoftheHDRimagesabilitiesintegratedina3Dprogram,theimagedeliversveryuseful,crisp,sharpandrealisticreflections,whichimprovesthefeelthatthesurroundingsarerealistic.

3.1.1.2 Perspective

In reality, different cameras are used to either generate photography or film andmovie production. In 3D, applications the camera is capable of both, making the

render process possible. In this research, we are concentrating on the photographyaspect. Even rendering from a view port is in fact working as a camera although

creatingacamerawillprovidetheuserwithvaluableparameterstocontrol.

Perspectiveisachievedthroughdifferentcameralenses,whereforexample50mmisthestandardlensfor35mmcameras.Changingthelenssizeleadstotheillusionthat

the cameras distance to the objects has changed. This includes distortions ofperspective. A 35 mm lens is called a lightwideangle. A 25 mm lens is a stronger

wideangle,whichstretchestheperspectiveevenmore,withtheresultthat itgetsabigger fisheye effect. A wideangle has a short focal distance, but a wider angle of

vision.Standing

in

front

of

abig

tower

for

example,

can

require

awide

angle

to

fit

the

wholebuildingintothepicture.An85mmlensiscalledlittletelephotolensorportrait

lens, because of its little loose of perspective. This results in a portrait with lowerdepth,whichmeansflatteningthefacea littlebit,wherethenose isnotstickingouttoomuch.Zoomingwith lenseswitha100135mmwilldecreasesthedepthandtheimageappearsmoreflat.Alargetelephotolensisabout200mm.Dependingonwhicheffectsarewanted, thecameradistanceto theobjectand/or the lenssizehas tobe

adjusted.

3.1.1.2.1 DepthintheimageDepthoffield

An image is always twodimensional, even if it was created in 3D. One of the majorproblems tostrugglewith is the illusion ofdepth in the image.Manyphotosappearflat and distracting. The whole trick is to catch the observers glimpse and lead itthroughtheimage.Thereareseveralwaystodothis,whereasthetoolsarealwaysthesame: the lines and forms in the image, their relations to each other and light andshadows. In the later case the photographer can for example chose a darkerforeground,whilethemainfocusliesinthebrighterbackground.Theobserverabsorbsintothepicture.Evenifthecontrastisnotthatextreme,thelightintensitydecreasesgettingnearertothepicturescorners.Thiscreatesaninvisibleborderthatkeepsthe

observerinsidethepicture.

The feeling ofdepth canalso be accomplished througha technique calledgradation(24). Ifanelement,adetailorapartofthescenepartlycoversorhidesanother,theobserver immediately realizes one object as near (in foreground) as the other (in

background)(Figure

10on

page

25).


25/52


Figure10:We immediatelyrecognize thattheblacksphere isnearest,since it is largestand in

frontoftheredandthewhiteone.

Thehumaneye istrainedtoseethingseveryday.Fromthebeginning,wehavebeenstudying objects and especially their size. This experience is important to estimatedistancesandspeedsintherealworldandinanelaboratedphotoaswell.Thelonger

thedistance

between

the

eye

and

the

object,

the

smaller

the

object

appears

to

be.

Biggerobjectsareoftenplacedintheforeground(24).Nevertheless,therelationshipbetweenobjectsismuchmoreimportantinthiscase.Atalltreerequiresacomparisontootherobjects,forexampleahumanbeingorahouse,sothatourexperiencecounts

astrueorrealistic.

An additional method to attain spatiality in the image is to work with thedepthof

focus inphotographyordepthoffield,as it iscalled in3D.Asmentionedbefore,thehumaneyecanonlyseesharpwiththecellsinthecenteroftheretina.However,sincethe eye is permanently moving around, our brain gives us the illusion that we seesharpthewholetime.Thedalliancebetweenthemovesistoosmall,lessthanwearerealizingthem.Butinphotography,thedepthoffocusgeneratesacenterofattentionintheimage,andleavesthemurkypartsasrespectively,fore andbackground.Thereare two types of depth of focus: short and wide. As itsname implies, a short depth

means that only a little part of the scene lies in the sharp region, while the rest ismurky. Usingawide depthof focus,hence the name, everything in the image lookssharp (24).This iscommon touse formore technicalpictures,as forexample IKEAs

Guideline images for their products database or the clean presentations in theircatalogue.

Depth of field in 3D works principally the same way, though it is very expensive in

rendering,inthesenseoftimeconsuming(Figure11).

Figure11:

An

image

rendered

with

Depth

of

Field

on,

with

the

focus

on

the

black

sphere.


26/52


3.1.1.3 In contrast to 3D

3.1.1.3.1 Noise,Softness

During the 19th

century, there started anewepoch in both literature and art that iscalled classicism. The artists were searching the truth mainly in nature. The Frenchartists and chemists LouisJacquesMand Daguerre and Joseph Nicphore Niepccreatedthefirstpermanentphotographybasedontheprinciplethatasilverandchalk

mixture darkens under exposure light, discovered by J. H. Schultz in the early 18th

century.Fromthismomentthephotographyreflectedtherealityinawayoilpaintings

nevercould.Ittookalongtimeuntilthecoloredfilmreallybrokethrough,wherethephotography felt as trustworthy as today. Nowadays, most of the huge photo

productionsforIKEAaremadedigitally.(24)

Producing a photo with a digital camera includes minor noise, while renderinggeneratestotallycrispyandcleanimages.Thisnoiseisgeneratedfromthecameras

change coupled device or color capture device (CCD) chip. This device uses lightsensitive material on a silicon chip to electronically detect the lights photons. This

incomingsignalistransferredfurtheralongarowofseparatepixels(pictureelements)

whereit

can

be

stored

as

color

information.

When

the

pixels

are

arranged

in

rows

and

columns,theassemblage iscalledtwodimensional.Acolor imageCCDsensorusesacheckerboardpatternofcolorfilters,representingthethreeprimarycolorsred,green

andblue.Alternatively,threeseparateCCDsensorsareapplied,onforeachcolor.(25)

3.1.1.3.2 Roundedcorners

Building object in 3D is similar to creating them in a mathematical correct way. A

polygonobject ismadeoffaceswithoutaphysicalthicknessas inreality.Evenathinsheet of paper has a depth. An equal problem appears on corners and edges. Real

objectscanhardlybeashardedgedastheyarein3D,whenstartingwiththecreation.The modeller has to be aware of those problems and correct them, at least for

photorealisticrendering.

3.1.2 Making the picture alive three main forces

3.1.2.1 Flaws in the perfected

Evenifthehumaneyeisphysicallyandbiologicallynotthebestorsophisticatedone,itcooperatesperfectlywiththebrain.Thisprovidesuswiththecapabilitytorecognizepeople,movements,pictures,andpatternsofeverytype.Particularly,when itcomestosymbolsandsignsthehumanbeing isquitefastandflexibletocommunicatewithother people, due to the evolution. On the other hand, in some areas, this exactlyability to recognize a pattern is not wanted, for instance in design and imagery.Recognition of patterns leads often to a feeling of boredom, although the observer

wantsto

experience

excitement,

suspense

and

harmony,

or

just

detect

athought

or

storybehindtheimage.

The secret of producing harmony do not lay in the construction of strait lines andanglesorinanimage,preciselydividedinequalpieces,butratherinafineinterruptionof lines and surfaces. Although the question about good, bad, beautiful andunaesthetic is very subjective, there are three main ideas to create a goodlookingimage:content,(clipping)detailandcomposition(24).

3.1.2.1.1 Content

The content should reflect the photographers meaning behind the image or whathe/shewantedustoseeorfeel,eveniftheobserverproduceshisowntruthaboutthe

picture.


27/52


3.1.2.1.1.1 ProppingProppingmeanstoplaceseveralobjects intoaroomsfurnitureorequipment,asforexample a couple of books in a bookshelf. Visiting IKEAs store or browsing the

catalogue, itbecomesclear,whypropping isveryuseful.Aboveall,abookshelffilledwithbooksandthingswillbebettersoldthananemptyone.Proppingmakesitmore

realisticlooking

and

personal;

it

provides

the

feeling

of

someone

actually

living

in

the

chosen environment. It can also inspire the customer how to combine their own

belongings.Aroomsetsproppingproducestheillusionofpeoplespresence,andcanalsotelltheviewermoreaboutwholivesthereandhowtheylive.Finedetailscanforexamplerevealwhetheritisafamilywithchildrenorayoungcouplewithpets.Thisisessentialforphotorealism,especiallyforIKEA.However,itstillhastobeaccomplishedwell and considered. Colors have to match to each other and all objects should be

placedharmonicallyandinthesenseofthecontent.Themeaningwithproppingistoillustrate the functionality or to emphasize the design with forms and colors (Figure

12). The number of propping elements depends on the scene, the light, thecompositionforall.Lessproppingoffersmoreownreflectionbytheobserver,though

the trick lies in finding a good balance inbetween the overcrowding and the

emptiness.

That

motto

for

minimalism

is

called

Less

is

more.

Figure12:Proppingpersonalizesaroom(26)

3.1.2.1.2 Clippingandformats,2832,55

Theborders

of

images

can

decide

as

well

how

the

image

feels

and

what

it.

By

clipping

offsomeessentialdetails,thecontentwillchange,too.

There are various formats, rectangular or quadratic. The first one can either be

landscape also called lyingformat, orportrait as a standingformat. As the nameimplies the landscape format appropriates for wide angled views, whereas smaller

subjects as a person, a portrait, comes out more harmonic in a smaller, but longerformat.Choosingaportraitformatforalandscapepicture,canconveyarestrictinganduncomfortable impression to the observer, standing in a room, watching through adoor(24).

Already 300 BC, the Greek mathematician Euclid described in one of his books the

godlyproportions,theGoldenRatio.Itcanbeexplainedasaline,dividedintwopieces

Aand

B

in

such

away,

that

A

is

to

B,

as

B

to

A+B

(Figure

13).


28/52


Figure13:TheGoldenRatio;AistoB,asBtothewholedistanceA+B

The Italian mathematician Leonardo Fibonacci (c. 11701250) found a series ofnumbers,theFibonaccinumbers,whichmaketheGoldenRatiomoreunderstandable:

1, 2, 3, 5, 8, 13, 21, 34, 55,89, 144, and so forth, building the algebraic sum of theprecursornumbers.

2000yearsago,theGreekarchitecturesbuildtheirtemples inthisperfectratio,as

well as many formats especially for papers and magazines are orientated to theseproportionsuntiltoday.Themostcommon formatforphotos is13x8cmasshown

below(Figure14).

Figure14:Thisformat13x8isacommonstandardforphotographs.

Thequadraticformatseemsasaverystaticandunchangeableone,andthereforeevenboring. It can be very exciting in combinations or for itself claiming an interestingcontentandcomposition.

3.1.2.1.3 CompositionImperfectmakesperfect

Nomatter

the

clipping

or

the

content,

an

image

needs

astructure

and

some

kind

of

considered composition, to be able to communicate with the observer. Thecompositionshallemphasizetheimagesmessageandthecontext,inwhichtheimage

isplaced.

Centering an object in the exact middle of the picture will generate an axialcomposition, which meanssymmetry, since thehumaneye often longs forcalmandorder. Placing the center of attention out of the middle of the picture, thephotographerproducesmoreexcitementandlife.Theasymmetryprofitsfromemptyspaces and their relation to the center of attention with different objects. Anasymmetricplacementofequipmentofferssometimesaneasiermethodtocatchthe

observersattention,sohefindsawayintotheimage.Itisalsopossibletoarrangethe

composition

by

the

Rule

of

Thirds

(24),

cutting

it

into

three

parts

and

placing

the

main

object of attraction in the left or in the right division. This method is similar toproportioningaccordingtotheGoldenratio.


29/52


Goodcompositionscanalsobecreatedwithcontrastsinsize,form,color,andsoforth.Thiskindofcompositionisabouttoemphasizecertainpartsintheimagewiththehelpof opposites and their effect to each other. This can accentuate for example a littledetailinthescene.Theruleistoavoidplacingtwoobjectsofthesamekindnexttoeachother,withthesamecolorvalue,saturation,orsizeandform(24).

Another

method

to

elicit

action

and

dramatic

from

an

image

is

to

construct

a

diagonallydirection with elements, surfaces and lines. The eyes automatically followthe lines into the depth of the picture, some kind ofcentered energy. However, thediagonal is also a very symbolic element, especially for Western people, writing andreadingfrom lefttoright.Whenthe line isstrivingupfromthebottom leftcornertothe upper right, it is a symbol of success, positive progress and opening, while the

oppositedirectionoftenmeansblockadeandprohibition.

The question is why does IKEA of Sweden want to place its furniture in a closedangle?BengtLarsson (2)providesaplausible,probableexplanation: It istotallyright

thatamaindiagonalgivesastraitdirectionandspeed toan image,but it isjustas

importantwhichobjectsareplaced in the scene.The image consistsofbalanceand

oftenseveralfieldsofattention,wherediagonal linesareusefulamongothers. IKEAs

Guidelinefurniture

is

always

placed

on

adiagonal,

falling

lines

with

an

angle

of

5

or

11,dependingontheobjectssize.Therefore,evenifthemaindirectionisadiagonal,theobjectoffersother lines thatarestrivingupagain.The imageof thesofa (Figure

15),thatBengthassentusmakesthis ideaclearer.Theobserverseyeswanderfromlefttothemiddleoftheimage,wheretheywillbestoppedandliftedupovertheback

support.Thislittle,shortbreak,whichtheeyestakeinthemiddleofthesofa,createapleasantfeelingofrestingandrelaxing,whichisdefinitelywanted.(2)

Figure15:BengtsexplainingimageoftheIKEAangle.Thesofaaboveispositionedintheright

angle,whilemirroredthesofabelowgeneratedunwantedfeelings,whenitcomestofurniture.


30/52


Whenmirroringtheimagehorizontal,thesofastandsontherisingdiagonal,andthenother unwanted byeffects will appear. First of all, the arm rest is blocking theobserver, while the eyes try to wander from left to the right. The sofa feels morerepelling,notinviting.Andwhilethefirstpieceoffurniturecatchestheobserversviewandreturnshimacomfortablefeeling,thesecondoneforceshimtoslideoutofthesofa, meaning he is not staying on the furniture. Details will be lost and the image

cannotcreate

the

wanted

pleasurable

feeling

of

rest.

(2)

Summarizing, it supplies one main factor to consider while creating a goodcomposition. An image without objects has no composition. That is why, thecompositor shall be aware of all visible objects in the scene and how their lines areworkingincombinationandrelationwitheachother.

3.1.2.1.4 Virtualphotorealism

There are many details in the real world that the observer sees unconsciously andtakesthemforgranted.Whiletheinteriordecoratorironsthingsasblankets,pillowsormattresses to get them creasefree, the 3Dartist struggles with creating accessoriesand fabric furniturewithseemsandstitches.This isachievedwith thehelpofbump

mapsor

displacement

maps.

The

photorealism

lies

somewhere

in

between,

depending

onmaterial, lightinganddistancefromthecamera(2).Materialsshall live:creasesin

fabricandcloth,natural irregularities insurfacessuchaswood!Nothing isdead leveland uniform. Breaking the perfectness makes it more acceptable for the eye and

thereforeevenmorecorrect,orphotorealistic.

Even the variation of textures is very essential for photorealism, since there are norepetitions of surfaces or texture in the real world. It facilitates the work of the 3Dartist,havingahugetexturelibraryathisdisposal.

3.2 Perception Psychology

Psychologyisthestudyofthementalprocessesandbehaviorofabrain.Perceptioninthe fieldofpsychology,on theotherhand, involveshowthebrain interpretsobjects

using the senses. In this research, only one of the senses will be investigated, thevision,whichistheabilitytodetectandunderstandelectromagneticwavesoflight.

3.2.1 Relative Brightness

A very strange and subjective question is how bright are things? A good example is

described in a book, written by Arnheim (27): It has often been observed that a

handkerchief atmidnight lookswhite, like a handkerchief at noon, although itmay

sendlesslighttotheeyesthanapieceofcharcoalunderthemiddaysun.Thisprovesthat brightness is a relative thing; it all depends on the distribution of light in the

environmentor

image

and

the

optical

psychological

process

of

the

eye

of

the

observer,

as well as the nervous system. This is also relative on the objects material ability toreflect light. Another example is if someone sits in a room at dawn and watches

television,focusestheireyesonthetelevisionscreen,andsomeonewalksintheroomlater on and states Why are the lights not turned on? flips the lightswitch, and it

becomesclearthatthesunhassetandtheenvironmentismuchdarkerwithoutevernoticing.Thisisalsothehumanmindworkingtogetherwiththepupilsoftheeyes,the

pupils grow wider to catch more light, as well as the mind compensating withunderstandingoftheenvironment.(28)

3.2.2 Color interpretation

The physical principles of how humans acquire and interpret the color in our

environmenthavealreadybeencoveredinthesecondchapter,butnotwhatthebraindoeswiththeinformationgathered.Aballthatcomesrollingacrossafieldcaneasybe

identified because of its movement, or its shape, maybe its texture but probably


31/52


because of its color, unless it is green, which would help it blend in with the greengrass. Althougha redball wouldclearlypaint aclear imageof colors importance forrecognition. With a few exceptions of physical defects and cultural and religiousbackground, everyone interprets color the same way. Even though almost everyonehasauniquetasteorsenseofcolor,weallhaveacommonunderstandingforallcolorsbecauseeveryonerelatestothesamething.Everyoneknowshowagreenappleorthe

colorapple

green

looks

like.

(28)

3.2.3 Shapes and forms

Shapes can be very subjective depending on the viewers earlier experiences andabilitytounderstandandrealize,veryoftentheuseofobjectssothattheshapecanbe

identified as a class of some sort, for example ananimal or a kitchen accessory. Forexample,ifapersonwenttobedlateatnight,whenitisdarkoutsideandmostofthelightsareout,andseesablackshapeonthebed.Ifthispersonlivesalone,thiscanbevery troubling,where ifyouhavea partner thathasgone tobedearlier, it isalmostexpected,andtheshapeisinstantlyrecognizedforwhatitis.

Or, if twodifferentpersons,oneveryproficient in kitchen equipment, theothernot

quiteas

skilled,

and

both

individuals

are

introduced

to

anew

sophisticated

type

of

blade sharpener. The first person might understand the use of the object instantly,whilethesecondonewouldnothaveaclueunlessexplainedtohim/her,thisstrictly

becauseoftheshapeoftheobject.(28)(29)

Figure16:Thisabstractschemeexplainshowthebrainandtheeyeworktogethertoperceivethe

environment.

3.2.3.1 Optical Illusions

Optical illusionsarestimuli that lie in theborder landsofwhat thehuman mindcan

comprehend. An optical illusion in most cases appears when the mind makes anassumptionthatiswrong,orwhenaconstantrecalibrationisforcedbythemind.Asin

the image above, there is aperfect example of how thebrain tries to help the eyesunderstand images (Figure 16). The brain reads them from left to right and takes

straight lines as an aid, as well as continuity. This precise problem is common in


32/52


computer graphics, in theory there is nothing wrong, but in the practical sense, theminddestroystheimage(Figure18).Therefore,itis important,nottoletcolorsburnoutoveralongrange,keepinginmindthattheeyecanswiftlyadjusttosuchburnoutareas and recognizedetails in them. Sowideregions with exact thesamecolor arenotphotorealistic(Figure17below).

Figure 17: The difference between physically and perceived intensity reveal how the eyes

deceivingus.

Figure18:ThesquareAhastheexactsamecolorassquareB,asthesamplesshow.Coloralways

dependsonitssurrounding.


33/52


4 PROCESS

During the firstweeks,bothofushadbeenresearchingthebackground informationfor the introduction part and the theoretical framework, in chapter 2 TheoreticalFramework.Wehavefoundsomeespecially interestingarticlesandevenbooksthatwerepartlyavailableontheinternet.

Then,wewenttoIKEAofCommunicationinlmhult,southernSweden,andlearnedagreatdealaboutthephotographyprocessandtheir3Dstudio.Thatinspiredusbothalotespeciallyforthebackgroundinformationforchapter3VisualResponse,whichis

the

additional

information

to

balance

the

theory

of

rendering.

The

Theoretical

Framework was more difficult to structure and make choices on what was most

importantforourresearch,andwhattoavoidkeepingthisresearchpaperfrombeingtoobroad.

The practical part of our thesis work consists of improving the old 3D image andgettingitmoresimilartotheoriginalphoto.

4.1 Study Excursion Description

Seeing and walking though IKEAs huge photo studio, we gained a great deal of

experience

and

an

insight

to

the

photographers

process.

The

problems,

which

photographersarestrugglingwith,areofcoursenotthesameasforthe3Dartists.Thephotostudio isbuiltasawidehallwith lotsofblackandwhitedraperies, thinwalls,

reflecting walls, and adjustable roofs. The remaining construction components foreachroom,equipmentandfurnishingvaries,dependingonwhichobject,furnitureor

room setting is going to be photographed. Bigger objects as sofas, beds, or roomsetupsdemandmoretime,space,andoftennaturallyamoreadvancedlightingsetup.

Thestudiohaslargewindowsaroundthebuildingswalls;thereforethephotographerhasthepossibilitytomakeuseoftheincomingdaylight;althoughhehastobeawareof that this light is changing during the day, which also can cause problems.Sometimes, theycover thewindowswithmatt, transparentplastic foil,orjustblack,isolatingdraperies.

Thephotographersworkindifferentways,someofthemworkontheirownandsomein larger teamsof twoormorepersons,dependingonhowhuge theobjectsand/or

scenesare.

4.2 Choosing a room

During our field trip, we asked Bengt Larsson for an original photograph of a roomsetting,whichwecanuseasreferenceforour3Dproduction.HefoundanexcellentimageofawhitekitchenthathasbeenprintedintheIKEAscataloguetwoyearsago.Thisavoidssecurityorpublishingissuesforourthesisresearchpaperaswell.


34/52


The immense advantage with this image of the kitchen is that IKEA had alreadymodeled, textured, and lit a whole scene in 3ds Max and rendered a quite goodlookingimage,asatestforpure3Dgeneratedroomsetups.Ourjobistoimprovethis3Dproductandachieveamorephotorealisticresult.

4.3 Image Analysis

4.3.1 Analyzing the original image

Beforestartingwithactualworkin3dsMax,boththeoriginalphoto(Figure1)andthe3d image (Figure 22) has to be analyzed. The main aspects are modeling, lighting,texturing,rendering,andremainingissuessuchascamerasettings.Wehavediscussed

whythe3Dpicturedoesnotfeelasphotorealisticastheoriginalimagedoes.

Figure19:Theoriginalimagewithalightanalysis


35/52


Figure20:Thecolorcompositionisimportanttogiveacertainfeelingtotheimage.Thiskitchen

feels quite clean because of the blackwhite contrast. The other colors attract the observers

attentionand

make

the

room

livelier.


36/52


Figure21:Proppingisimportanttoshowwhoislivingintheroomandhowitisused.


37/52


Figure22:Thisisthe3Dimageoftheoriginalphoto,modeled,textured,lightenedandrendered

byLindaFerroniatIKEA.

4.3.2 Things to change in the 3D scene (Figure 22)

4.3.2.1 Modeling

Ourmaingoal istoachieveaphotorealisticsuperiorlooking image.Sincethere isanoriginalphototoreferto,the3Dkitchenhastobeasidenticalaspossible.Thisiswhy

evensomemodelingchangeswouldhavetoberealizedinthebeginningprocess.

Starting from the top to the bottom (Figure 22), some of the glasses positions havebeen moved around, to avoid toomuch ofa straight order in the glass cabinet.The

doorhandlesseemed toobigandhad tobemovedupslightly.Themetalbarundertheglasscabinetwasmoveddownandscaledup,aswellasthemountingequipment.

Thewatertapwasrotatedseveraldegreestotheleft,thedishsoapsbottlewasscaleddownonethirdandoneofthedishbrusheswasrotatedtoo,wheretheydonothang

onthebarintheexactsameway.

Bothceilinglamps,hangingoverthetable,weretoolargeandhadtobescaleddown

somecentimeters.Agreatdealofremodelingworkwasneededontheairexhausterovertheoven,scalingupthelampsandgivingsomedepthtotheunderside.


38/52


Theovenseemsveryflat, therefore itwasrepositionedslightlyand theovendoor isnow open, creating a natural, narrow gap. The same phenomenon occurs on thekitchendoors, too,wherethegapsweretoosmall.Wechangedthebumpmapwiththemodeled,realgeometry.

4.3.2.2

LightingThe lighting isveryessential for thewholeappearanceofcolors,shadows,andeven

themodels.Afteranalyzingthe3D image,somechangeshadtobeachieved.Firstofall,amainlightismissing,whichlightsthekitchencabinetinthefrontofthepicture.Thislampseemstobequitebright,maybeaswhiteasdaylight.

The second daylight source from the window has been changed; its light focus is

movedmoretothemiddleofthekitcheninsteadofilluminatingonlytherightside.Toachieve the original photos brightness and feeling of openness, this light is the keylight.

The floor is build of dark brown wood boards, somewhat reflecting. That requiresplenty of light, where itdoes not appear as a homogenyblack floor. Threedifferent

coldlights

are

blending

together

on

the

floor:

from

the

window,

the

front,

and

the

secondroomintheleftoftheroomattheback.

Theceilinglampsoverthetablearemoreyellowishthaninthe3Dimage.Toobtainabalancebetweenwarmandcoldlightinthescene(asintheoriginalphoto,Figure19),the ceiling lamps can be opposite pole to the incoming daylight from the windows.Theselampsarealreadyilluminatingwarmerlight;howeverthelampitselfshallshinemoreyellowish,notwhite.Thiscanbepartofthetexturingprocess.Whenthetexture

ofthelampsinsidewasfinishedandlookedgood,thereappearedotherartifacts,suchasyellowdotsallovertheimagefromthelampreflections.Therefore,wedecidedto

achievearealisticlightinginthewholesceneandeventuallyrendertheyellowlampsseparatelyandpostthemafterwards.

All

metallic

materials

appear

very

lifeless

and

dull.

When

the

lighting

is

adjusted,

it

can

stillrequireaHDRIinthetexture.

Overall,wefindthatthekitcheninthe3Dimagewastoodarkandhazy.Animprovedlightingisessentialtoaccomplishtheoriginalimageslightandopenfeeling.

4.3.2.3 Texturing

The texturing process is often very timeconsuming, since every material has to be

adjusted to the lighting. Settings like color, reflectivity or displacement maps oftenneedvarioustweaks.

As mentioned before, the metallic materials have to be more realistic, reflecting aHDRIforexample.Thekitchendoorsarealsorathertoomatt,whiletherealdoorsarevarnished.

The

glass

tops

in

the

glass

cabinet

are

too

green,

and

should

have

amore

bluishtexture.Detailssuchasfixinga labelforthetransparentbottlebythesinkcanaccomplishmorephotorealism,becauseitismoretrustworthywithrealisticdetails.

The oranges do not appear very delicious. The texture should be more orange thanyellow,aswellasthefruitshouldberotatedabit,whereitreflectsthelightfromtheceiling lamps. They need to feel fresh and kind of shiny. Photographing food is ascienceofitsown.

The potholders are more similar to towel fabric. Since the camera focus lies in thebackground,detailsofthepotholdersareblurred.Towelmaterialwillprovideasoftertouchtotheimage.

In

the

original

image,

it

is

obvious

that

the

wooden

floor

consists

of

boards,

while

the

3Dfloorisflat,nothavinganygapsinbetween.Thislackofmodelingcanbearranged

withthehelpofadisplacementmap.


39/52


4.3.2.4 Rendering

Thefinalimagewasrenderedwiththefollowingrendersettings:

GlobalSwitches:Nohiddenlights,nodefaultlights,

AntiAliasing:AdaptiveQMC,CatmullRom

AdaptiveQMC

image

sampler:

min

subdiv

1,

max

subdiv

2

IndirectIllumination(GI):GIcausticsrefractive,primarybounces:QuasiMonteCarlo,secondarybounces:Lightcache.

Lightcache:Subdivs:500800forweb,1500forhighqualityimages(3000px),numberofpasses=howmanyprocessorsareavailable

Environment:skylighton(colorwhite),reflection/refractionon

rQMCSampler:Noisethreshold:0.0005,Minsamples:16,GlobalsubdivsMultiplier:3

4.3.2.5 Problems

Mostproblematic

are

our

insufficient

computer

capacity.

The

scene

turned

out

very

complexandtimeconsumingtorenderonmycomputerortheo

Date post:	08-Aug-2018
Category:	Documents
Upload:	juan-patricio-campos-torres
View:	231 times
Download:	0 times

render fotorealista

Documents