44
18/07/02 1 Min H. Kim KAIST CS580 Computer Graphics CS580: Computer Graphics Min H. Kim KAIST School of Computing Min H. Kim KAIST CS580 Computer Graphics Elements of Computer Graphics 2 Rendering Material model Geometry Light Virtual photography

Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

  • Upload
    others

  • View
    5

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

1

MinH.Kim KAISTCS580ComputerGraphics

CS580:ComputerGraphics

MinH.KimKAISTSchoolofComputing

MinH.Kim KAISTCS580ComputerGraphics

ElementsofComputerGraphics

2Rendering

MaterialmodelGeometry Light

Virtualphotography

Page 2: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

2

MinH.Kim KAISTCS580ComputerGraphics

HDRENVIRONMENTMAP

3

MinH.Kim KAISTCS580ComputerGraphics

CapturingEnvironmentMaps•  Photographingalightprobeproducesanenvironmentmaprepresentingincidentradiancefromalldirections.

4 B(w) = 2( w ⋅ n)n − w

Page 3: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

3

MinH.Kim KAISTCS580ComputerGraphics

CapturingEnvironmentMaps

5

•  Howtoremovethecamerafromtheenvironmentmap?

MinH.Kim KAISTCS580ComputerGraphics

SphericalEnvironmentMap&CubeMap•  Galileo’sTomb

6

http://www.pauldebevec.com/Probes/

Page 4: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

4

MinH.Kim KAISTCS580ComputerGraphics

Environmentcubemaps•  Texturescanalsobeusedtomodeltheenvironmentinthedistancearoundtheobjectbeingrendered.

•  Inthiscase,wetypicallyuse6squaretexturesrepresentingthefacesofalargecubesurroundingthescene.

7

MinH.Kim KAISTCS580ComputerGraphics

Environmentcubemaps•  Eachtexturepixelrepresentsthecolorasseenalongonedirectionintheenvironment.

•  Thisiscalledacubemap.GLSLprovidesacube-texturedatatype,samplerCubespecificallyforthispurpose.

8

Page 5: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

5

MinH.Kim KAISTCS580ComputerGraphics

Environmentcubemaps•  Duringtheshadingofapoint,wecantreatthematerialatthatpointasaperfectmirrorandfetchtheenvironmentdatafromtheappropriateincomingdirection.

9

MinH.Kim KAISTCS580ComputerGraphics

Environmentmapshader•  Wecalculateinthepreviouslecture.•  Thisbouncedvectorwillpointpointstowardstheenvironmentdirection,whichwouldbeobservedinamirroredsurface.

•  Bylookingupthecubemap,usingthisdirection,wegivethesurfacetheappearanceofamirror.

10

B(v)

Page 6: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

6

MinH.Kim KAISTCS580ComputerGraphics

Environmentmapshader•  Fragmentshader

11

#version130uniformsampler2DuTexUnit0;invec3nNormal;invec4vPosition;outvec4fragColor;vec3reflect(vec3w,vec3n){returnn*(dot(w,n)*2.0)-w;//bouncevector}voidmain(){vec3normal=normalize(vNormal);vec3reflected=reflect(normalize(vec3(-vPosition)),normal);vec4texColor0=textureCube(uTexUnit0,reflected);fragColor=vec4(texColor0.r,texColor0.g,texColor0.b,1.0);;}

[0,0,0,1]t

MinH.Kim KAISTCS580ComputerGraphics

Environmentmapshader•  -vPositionrepresentstheviewvector•  textureCubeisaspecialGLSLfunctionthattakesadirectionvectorandreturnsthecolorstoredatthisdirectioninthecubetexturemap.

•  Hereweassumeeye-coordinates,butframechangesmaybeneeded.

12

v

Page 7: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

7

MinH.Kim KAISTCS580ComputerGraphics

Environmentmapshader•  Thiscanbeusedforrefraction.

13

MinH.Kim KAISTCS580ComputerGraphics

HIGH-DYNAMIC-RANGEIMAGING

MeasureRadianceas2DImages

14

Acknowledgements:someofslidesarecourtesyofProf.PaulDebevec(USC),Prof.AlexeiEfros(CMU),andProf.FredoDurand(MIT)

Page 8: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

8

MinH.Kim KAISTCS580ComputerGraphics

TheProblem

•  Limitofdigitalcamera

15

MinH.Kim KAISTCS580ComputerGraphics

Problem:DynamicRange

•  Therealworldishighdynamicrange

16

1 1,500

2,000,000,000400,000

25,000

Page 9: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

9

MinH.Kim KAISTCS580ComputerGraphics

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

Multipleexposurephotography

17CourtesyofF.Durand

MinH.Kim KAISTCS580ComputerGraphics

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

Multipleexposurephotography

18CourtesyofF.Durand

Page 10: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

10

MinH.Kim KAISTCS580ComputerGraphics

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

Multipleexposurephotography

19CourtesyofF.Durand

MinH.Kim KAISTCS580ComputerGraphics

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

Multipleexposurephotography

20CourtesyofF.Durand

Page 11: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

11

MinH.Kim KAISTCS580ComputerGraphics

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

Multipleexposurephotography

21CourtesyofF.Durand

MinH.Kim KAISTCS580ComputerGraphics

Howdowevaryexposure?•  Options:– Shutterspeed– Aperture–  ISO– Neutraldensityfilter

22

Page 12: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

12

MinH.Kim KAISTCS580ComputerGraphics

Tradeoffs•  Shutterspeed–  Range:~30secto1/4000sec(6ordersofmagnitude)–  Pros:reliable,linear–  Cons:sometimesnoiseforlongexposure

•  Aperture–  Range:~f/1.4tof/22(2.5ordersofmagnitude)–  Cons:changesdepthoffield–  Usefulwhendesperate

23

MinH.Kim KAISTCS580ComputerGraphics

Tradeoffs•  ISO–  Range:~100to1600(1.5ordersofmagnitude)

–  Cons:noise–  Usefulwhendesperate

•  Neutraldensityfilter–  Range:upto4densities(4ordersofmagnitude)&canbestacked

–  Cons:notperfectlyneutral(colorshift),–  notveryprecise,needtotouchcamera(shake)–  Pros:workswithstrobe/flash,goodcomplementwhendesperate

24

Page 13: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

13

MinH.Kim KAISTCS580ComputerGraphics

HDRimageusingmultipleexposure

•  GivenNphotosatdifferentexposure•  RecoveranHDRcolorforeachpixel

25

MinH.Kim KAISTCS580ComputerGraphics

Ifweknowtheresponsecurve

•  Justlookuptheinverseoftheresponsecurve•  Buthowdowegetthecurve?

26

Pixel value

scene value

Page 14: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

14

MinH.Kim KAISTCS580ComputerGraphics

Calibratingtheresponsecurve•  Twobasicsolutions–  Varysceneluminanceandseepixelvalues

•  Assumeswecontrolandknowsceneluminance

–  Varyexposureandseepixelvalueforonesceneluminance•  Butnotethatwecanusuallynotvaryexposuremorefinelythanby1/3stop

•  Bestofboth:–  Varyexposure–  Exploitthelargenumberofpixels

27

MinH.Kim KAISTCS580ComputerGraphics

TheAlgorithm•  Imageseries

28

• 3

• 1 •

2

t = 1/100 sec

• 3

• 1 •

2

t = 1 sec

• 3

• 1 • 2

t = 1/1000 sec

• 3

• 1 •

2

t = 10 sec

• 3

• 1 •

2

t = 1/10 sec

Z = f (H )H = E ⋅ Δtlog(H )= log(E)+ log(Δt)where Z is pixel value, H is exposure,E is radiance, Δt is exposure time.

CourtesyofPaulDebevec

Page 15: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

15

MinH.Kim KAISTCS580ComputerGraphics

Responsecurve

•  Exposureisunknown,fittofindasmoothcurve

29log Exposure

curve

Pixe

l val

ue 3

1

2

log Exposure

Pixe

l val

ue

Assumingunitradianceforeachpixel

Afteradjustingradiancestoobtainasmoothresponse

MinH.Kim KAISTCS580ComputerGraphics

TheMath

•  Letg(z)bethediscretelogarithmicinverseresponsefunction

•  Foreachpixelsiteiineachimagej,want:

30g(Zij )= log(Ei )+ log(Δt j )

Z = f (H ) → logH = log f −1(Z ) → logH = g(Z )

Z = f (H )H = E ⋅ Δtlog(H )= log(E)+ log(Δt)where Z is pixel value, H is exposure,E is radiance, Δt is exposure time.

Page 16: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

16

MinH.Kim KAISTCS580ComputerGraphics

TheMath•  Solvetheoverdeterminedlinearsystem:

31

log(Ei )+ log(Δt j )− g(Zij )⎡⎣ ⎤⎦2

j=1

P

∑ +λ g ''(z)2z=Zmin

Zmax

∑i=1

N

Dataterm Regularizationterm

MinH.Kim KAISTCS580ComputerGraphics

TheMath•  Howtoreconstructradiance:

•  Weightingfactorhandlesthenoisearoundjointsofpieces

32

logEi =

w(Zij ) g(Zij )− log(Δt j )( )j=1

P

w(Zij )j=1

P

Page 17: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

17

MinH.Kim KAISTCS580ComputerGraphics

Matlabcodefunction[g,lE]=gsolve(Z,B,l,w) %Zispixelvalue,Bisexposurefactorn=256;A=zeros(size(Z,1)*size(Z,2)+n+1,n+size(Z,1));b=zeros(size(A,1),1);k=1; %Includethedata-fittingequationsfori=1:size(Z,1)forj=1:size(Z,2)wij=w(Z(i,j)+1);A(k,Z(i,j)+1)=wij;A(k,n+i)=-wij;b(k,1)=wij*B(i,j);k=k+1;endendA(k,129)=1; %Fixthecurvebysettingitsmiddlevalueto0k=k+1;fori=1:n-2 %IncludethesmoothnessequationA(k,i)=l*w(i+1);A(k,i+1)=-2*l*w(i+1);A(k,i+2)=l*w(i+1);k=k+1;Endx=A\b; %SolvethesystemusingSVDg=x(1:n);lE=x(n+1:size(x,1));

33

CourtesyofPaulDebevec

MinH.Kim KAISTCS580ComputerGraphics

Result:digitalcamera•  Reconstructedcameraresponse

34logExposure

Pixelvalue

RecoveredresponsecurveKodakDCS4601/30to30sec

Page 18: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

18

MinH.Kim KAISTCS580ComputerGraphics

Reconstructedradiancemap

35

CourtesyofP

aulD

ebev

ec

MinH.Kim KAISTCS580ComputerGraphics

Result:colorfilm•  KodakGold,ASA100

36

Page 19: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

19

MinH.Kim KAISTCS580ComputerGraphics

Recoveredresponsecurves

37

Red Green

Blue RGB

MinH.Kim KAISTCS580ComputerGraphics

TheRadianceMap•  Measuresceneradiancebyusingdigitalcamera

38

Page 20: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

20

MinH.Kim KAISTCS580ComputerGraphics

TheRadianceMap

39

Linearlyscaledtodisplaydevice

MinH.Kim KAISTCS580ComputerGraphics

HDRimageprocessing

•  Importantalsofordepthoffieldpost-process

40

Motionblurappliedtolow-dynamic-rangepicture

Motionblurappliedtohigh-dynamic-rangepicture

Realmotion-blurredpicture

ImagesfromDebevec&Malik1997

Page 21: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

21

MinH.Kim KAISTCS580ComputerGraphics

SmarterHDRcapture•  Automaticexposurealignment•  Ghostremoval•  Lensflareremoval•  ImplementingHDRincameras

41

ImagesGregWard

MinH.Kim KAISTCS580ComputerGraphics

Imageregistration

•  Howtorobustlycompareimagesofdifferentexposure?

•  Useablackandwhiteversionoftheimagethresholdedatthemedian– Median-ThresholdBitmap(MTB)

•  Findthetranslationthatminimizesdifference

•  Accelerateusingpyramid

42

Page 22: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

22

MinH.Kim KAISTCS580ComputerGraphics

Alignmentresult

43

MinH.Kim KAISTCS580ComputerGraphics

HDRVideo

•  GenerateHDRvideobyrapidlyvaryingtheexposureofeachframe(automaticexposurecontrolandstitchingneighboringframes)

44

Kangetal.SIG2003

Page 23: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

23

MinH.Kim KAISTCS580ComputerGraphics

Hardwaresolutions

45

MinH.Kim KAISTCS580ComputerGraphics

HDRImagingHardware

•  Mosaicneutral-densityfilterforspatialvaryingexposureimaging.Fourdifferentexposuresofneutraldensityfilterareinstalledinfrontofthedetectorarray.Thedifferencebetweenneutraldensityise3=4e2=16e1=64e0.NayarandMitsunaga[2000].

46

Page 24: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

24

MinH.Kim KAISTCS580ComputerGraphics

Fileformats•  PortableFloatMap(.pfm)– 4bytes(4x8=32bits)perchannel– 12bytesperpixel(=4bytesx3ch.)– TextheadersimilartoJeffPoskanzer’s.ppmimageformat(FloatingPointTIFFsimilar):

47

MinH.Kim KAISTCS580ComputerGraphics

Fileformats

•  Radianceformat(.pic,.hdr)– 4bytes(4x8=32bits)perpixel–  (145,215,87,149)=(145,215,87)*2^(149-128)(1190000,1760000,713000)

–  (145,215,87,103)=(145,215,87)*2^(103-128)(0.00000432,0.00000641,0.00000259)

48Ward,Greg."RealPixels,"inGraphicsGemsIV,editedbyJamesArvo,AcademicPress,1994

Page 25: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

25

MinH.Kim KAISTCS580ComputerGraphics

Fileformats

•  ILM’sOpenEXR(.exr)– 2bytes(2x8=16bits)perchannel(2x3=6bytes(16x3bits)perpixel)

– Losslesscompressionsupport– Half-precisionfloat– 65504(maxhalfprecision)– http://www.openexr.net/– Multi-channelsupported

49

MinH.Kim KAISTCS580ComputerGraphics

SunnybrookHDRdisplay•  Usebrightsource+two8-bitmodulators•  Transmissionmultipliestogether•  Over10,000:1dynamicrangepossible

50

Page 26: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

26

MinH.Kim KAISTCS580ComputerGraphics

HighDynamicRangeDisplay

51

MinH.Kim KAISTCS580ComputerGraphics

Howitworks

52

Page 27: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

27

MinH.Kim KAISTCS580ComputerGraphics

WhatifedgecontrastexceedsLCDrange?•  Observerscannottellwhenthishappensbecausetheeyehaslimitedlocalcontrastcapacityduetoscattering

53

MinH.Kim KAISTCS580ComputerGraphics

HDRDisplay

54

BrightSideDR37-P(nowDolby)

Page 28: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

28

MinH.Kim KAISTCS580ComputerGraphics

Displaytheinformation•  Matchlimitedcontrastofthemedium•  Preservedetails

55

10-6 106

10-6 106

Real world

Picture

Low contrast

High dynamic range

MinH.Kim KAISTCS580ComputerGraphics

Insightoftonemapping

•  TumblinandRushmeier(1993)’stone-reproductionoperatorcomprisesreal-worldobservations,inversedisplayobservations,andaninversedisplaydevicefunctionthatachievesaperceptualmatchbetweenreal-worldobservationandtheobservationofthereproducedimageonthedisplay.

56

Page 29: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

29

MinH.Kim KAISTCS580ComputerGraphics

Howhumansdealwithdynamicrange?

•  We'resensitivetocontrast(multiplicative)–  Aratioof1:2isperceivedasthesamecontrastasaratioof100to200

– Makessensebecauseilluminationhasamultiplicativeeffect–  Usethelogdomainasmuchaspossible

•  Dynamicadaptation(verylocalinretina)–  Pupil–  Neural/chemical

•  Differentsensitivitytospatialfrequencies

57

MinH.Kim KAISTCS580ComputerGraphics

ContrastSensitivity•  SineWavegrating•  Whatcontrastisnecessarytomakethegratingvisible?

58

Page 30: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

30

MinH.Kim KAISTCS580ComputerGraphics

ContrastSensitivityFunction(CSF)

•  CampbellRobsoncontrast-sensitivitychart

59

loga

rithm

icdecreaseofcon

trast

logarithmicincreaseofspatialfrequency

MinH.Kim KAISTCS580ComputerGraphics

ContrastSensitivityFunction(CSF)

•  Lowsensitivitytolowfrequencies

•  Importanceofmediumtohighfrequencies

•  Mostmethodstodealwithdynamicrange,reducingthecontrastoflowfrequencies

•  Butkeepthecolor

60

Page 31: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

31

MinH.Kim KAISTCS580ComputerGraphics

Contrastreduction•  Input:high-dynamic-rangeimage– Floatingpointperpixel

61 CourtesyofF

redo

Duran

d

MinH.Kim KAISTCS580ComputerGraphics

Naïvetechnique•  Scenehas1:10,000contrast,displayhas1:100•  Simplestcontrastreduction?

62 CourtesyofF

redo

Duran

d

Page 32: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

32

MinH.Kim KAISTCS580ComputerGraphics

Naïve:Gammacompression

•  XàXϒ(whereϒ=0.45inthiscase)•  But...colorsarewashed-out.Why?

63 CourtesyofF

redo

Duran

d

MinH.Kim KAISTCS580ComputerGraphics

Gammacompressiononintensity

•  ColorsareOK,butdetails(intensityhigh-frequency)areblurred

64 CourtesyofF

redo

Duran

d

Page 33: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

33

MinH.Kim KAISTCS580ComputerGraphics

Oppenheim1968,Chiuetal.1993

•  Reducecontrastoflow-frequencies•  Keephighfrequencies

65 CourtesyofF

redo

Duran

d

MinH.Kim KAISTCS580ComputerGraphics

Thehaloartifact

•  Forstrongedges•  Becausetheycontainhighfrequency

66 CourtesyofF

redo

Duran

d

Page 34: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

34

MinH.Kim KAISTCS580ComputerGraphics

Whyhaloartifacthappen

67

MinH.Kim KAISTCS580ComputerGraphics

DurandBilateralFiltering•  Donotbluracrossedges•  Non-linearfiltering

68 CourtesyofF

redo

Duran

d

Page 35: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

35

MinH.Kim KAISTCS580ComputerGraphics

StartwithGaussianfiltering

69CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

StartwithGaussianfiltering

70CourtesyofFredoDurand

Page 36: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

36

MinH.Kim KAISTCS580ComputerGraphics

StartwithGaussianfiltering

71CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

Gaussianfilterasweightedaverage

72CourtesyofFredoDurand

Page 37: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

37

MinH.Kim KAISTCS580ComputerGraphics

Theproblemofedges

73CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

PrincipleofBilateralfiltering

74CourtesyofFredoDurand

Page 38: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

38

MinH.Kim KAISTCS580ComputerGraphics

Bilateralfiltering

75CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

Bilateralfiltering

76CourtesyofFredoDurand

Page 39: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

39

MinH.Kim KAISTCS580ComputerGraphics

Normalizationfactor

77CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

Bilateralfilteringisnon-linear

78CourtesyofFredoDurand

Page 40: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

40

MinH.Kim KAISTCS580ComputerGraphics

OtherView

79CourtesyofFredoDurand

MinH.Kim KAISTCS580ComputerGraphics

Handlinguncertainty

80CourtesyofFredoDurand

Page 41: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

41

MinH.Kim KAISTCS580ComputerGraphics

DurandandDorsey(SIG2002)•  FollowingOppenheimetal,fromtheinputimageand

illuminanceimage,thereflectanceimagecouldbereconstructedbydividingtheinputandilluminanceimage.

•  Thesmoothilluminationiscalled“baselayer”,whereastheresultofthedivisioniscalled“detaillayer”

•  CompressionisdoneinthelogarithmicchannelsoftheluminancechannelintheYxycolorspace

81

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

82DurandandDorsey(SIG2002)

Page 42: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

42

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

83DurandandDorsey(SIG2002)

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

84DurandandDorsey(SIG2002)

Page 43: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

43

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

85DurandandDorsey(SIG2002)

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

86DurandandDorsey(SIG2002)

Page 44: Elements of Computer Graphics - KAISTvclab.kaist.ac.kr/cs580/slide13-HDR.pdf · 2018-07-02 · 18/07/02 5 Min H. Kim KAIST CS580 Computer Graphics Environment cube maps • During

18/07/02

44

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

87DurandandDorsey(SIG2002)

MinH.Kim KAISTCS580ComputerGraphics

Durand’scontrastreduction

88DurandandDorsey(SIG2002)