Transcript
Page 1: Perception and Measurement of Light, Color, and Appearance

Perception and Measurement Perception and Measurement ofof

Light, Color, and AppearanceLight, Color, and Appearance

Page 2: Perception and Measurement of Light, Color, and Appearance

ProblemsProblems

• How do cameras measure light and How do cameras measure light and color?color?– RadiometryRadiometry

• How do humans perceive light and How do humans perceive light and color?color?– PhotometryPhotometry

• How do monitors display light and How do monitors display light and color?color?

Page 3: Perception and Measurement of Light, Color, and Appearance

IntensityIntensity

• Perception of intensity is nonlinearPerception of intensity is nonlinear

Amount of lightAmount of light

PerceivedPerceivedbrightnessbrightness

Page 4: Perception and Measurement of Light, Color, and Appearance

Modeling Nonlinear Intensity Modeling Nonlinear Intensity ResponseResponse

• Brightness (Brightness (BB) usually modeled as a ) usually modeled as a logarithm or power law of intensity (logarithm or power law of intensity (II))

• Exact curve varies with ambient light,Exact curve varies with ambient light,adaptation of eyeadaptation of eye

3/1

log

IB

IkB

3/1

log

IB

IkB

II

BB

Page 5: Perception and Measurement of Light, Color, and Appearance

CRT ResponseCRT Response

• Power law for Intensity (Power law for Intensity (II) vs.) vs.applied voltage (applied voltage (VV))

• Other displays (e.g. LCDs) contain Other displays (e.g. LCDs) contain electronics to emulate this lawelectronics to emulate this law

5.2

VI

5.2

VI

Page 6: Perception and Measurement of Light, Color, and Appearance

Digression: Monitor KnobsDigression: Monitor Knobs

• ““Brightness” knob is offsetBrightness” knob is offset

• ““Contrast” knob is scaleContrast” knob is scale

• Yes, the names Yes, the names areare misleading… misleading…

)( brightnessVcontrastI )( brightnessVcontrastI

Page 7: Perception and Measurement of Light, Color, and Appearance

CamerasCameras

• Original cameras based on Vidicon obey Original cameras based on Vidicon obey power law for Voltage (V) vs. Intensity power law for Voltage (V) vs. Intensity (I):(I):

• Vidicon + CRT = almost linear!Vidicon + CRT = almost linear!

45.0

IV

45.0

IV

Page 8: Perception and Measurement of Light, Color, and Appearance

CCD CamerasCCD Cameras

• Camera gamma codified in NTSC Camera gamma codified in NTSC standardstandard

• CCDs have linear response to incident CCDs have linear response to incident lightlight

• Electronics to apply required power lawElectronics to apply required power law

• So, pictures from most cameras So, pictures from most cameras (including digital still cameras) will have (including digital still cameras) will have = 0.45= 0.45

Page 9: Perception and Measurement of Light, Color, and Appearance

Consequences for VisionConsequences for Vision

• Output of most cameras is not linearOutput of most cameras is not linear

• Know what it is! (Sometimes system Know what it is! (Sometimes system automagically applies “gamma correction”)automagically applies “gamma correction”)

• Necessary to correct raw pixel values for:Necessary to correct raw pixel values for:– Reflectance measurementsReflectance measurements

– Shape from shadingShape from shading

– Photometric stereoPhotometric stereo

– Recognition under variable lightingRecognition under variable lighting

Page 10: Perception and Measurement of Light, Color, and Appearance

Consequences for VisionConsequences for Vision

• What about e.g. edge detection?What about e.g. edge detection?– Often want “perceptually significant” edgesOften want “perceptually significant” edges

– Standard nonlinear signal close to (inverse Standard nonlinear signal close to (inverse of)of)human responsehuman response

– Using nonlinear signal often the “right Using nonlinear signal often the “right thing”thing”

Page 11: Perception and Measurement of Light, Color, and Appearance

Contrast SensitivityContrast Sensitivity

• Contrast sensitivity for humans about Contrast sensitivity for humans about 1%1%

• 8-bit image (barely) adequate if using 8-bit image (barely) adequate if using perceptual (nonlinear) mappingperceptual (nonlinear) mapping

• Frequency dependent: contrast Frequency dependent: contrast sensitivity lower for high and very low sensitivity lower for high and very low frequenciesfrequencies

Page 12: Perception and Measurement of Light, Color, and Appearance

Contrast SensitivityContrast Sensitivity

• Campbell-Robson contrast sensitivity Campbell-Robson contrast sensitivity chartchart

Page 13: Perception and Measurement of Light, Color, and Appearance

Bits per Pixel – Scanned PicturesBits per Pixel – Scanned Pictures

8 bits / pixel / color8 bits / pixel / color 6 bits / pixel / color6 bits / pixel / color

Marc Levoy / Hanna-BarberaMarc Levoy / Hanna-Barbera

Page 14: Perception and Measurement of Light, Color, and Appearance

Bits per Pixel – Scanned Pictures Bits per Pixel – Scanned Pictures (cont.)(cont.)

5 bits / pixel / color5 bits / pixel / color 4 bits / pixel / color4 bits / pixel / color

Marc Levoy / Hanna-BarberaMarc Levoy / Hanna-Barbera

Page 15: Perception and Measurement of Light, Color, and Appearance

Bits per Pixel – Line DrawingsBits per Pixel – Line Drawings

8 bits / pixel / color8 bits / pixel / color 4 bits / pixel / color4 bits / pixel / color

Marc Levoy / Hanna-BarberaMarc Levoy / Hanna-Barbera

Page 16: Perception and Measurement of Light, Color, and Appearance

Bits per Pixel – Line Drawings Bits per Pixel – Line Drawings (cont.)(cont.)

3 bits / pixel / color3 bits / pixel / color 2 bits / pixel / color2 bits / pixel / color

Marc Levoy / Hanna-BarberaMarc Levoy / Hanna-Barbera

Page 17: Perception and Measurement of Light, Color, and Appearance

ColorColor

• Two types of receptors: rods and Two types of receptors: rods and conescones

Rods and conesRods and cones Cones in foveaCones in fovea

Page 18: Perception and Measurement of Light, Color, and Appearance

Rods and ConesRods and Cones

• RodsRods– More sensitive in low light: “scotopic” visionMore sensitive in low light: “scotopic” vision

– More dense near peripheryMore dense near periphery

• ConesCones– Only function with higher light levels:Only function with higher light levels:

“photopic” vision“photopic” vision

– Densely packed at center of eye: foveaDensely packed at center of eye: fovea

– Different types of cones Different types of cones color vision color vision

Page 19: Perception and Measurement of Light, Color, and Appearance

ColorColor

• 3 types of cones: L, M, S3 types of cones: L, M, S

Page 20: Perception and Measurement of Light, Color, and Appearance

Tristimulus ColorTristimulus Color

• Any distribution of light can be summarized Any distribution of light can be summarized by its effect on 3 types of conesby its effect on 3 types of cones

• Therefore, human perception of color is aTherefore, human perception of color is a3-dimensional space3-dimensional space

• MetamerismMetamerism: different spectra, same : different spectra, same responseresponse

• Color blindness: fewer than 3 types of conesColor blindness: fewer than 3 types of cones– Most commonly L cone = M coneMost commonly L cone = M cone

Page 21: Perception and Measurement of Light, Color, and Appearance

ColorspacesColorspaces

• Different ways of parameterizing 3D Different ways of parameterizing 3D spacespace

• RGBRGB– Official standard: R = 645.16 nm,Official standard: R = 645.16 nm,

G = 526.32 nm, B = 444.44 nmG = 526.32 nm, B = 444.44 nm

– Most monitors are some approximation to Most monitors are some approximation to thisthis

Page 22: Perception and Measurement of Light, Color, and Appearance

XYZ ColorspaceXYZ Colorspace

• RGB can’t represent all pure RGB can’t represent all pure wavelengths with positive valueswavelengths with positive values– Saturated greens would require negative Saturated greens would require negative

redred

• XYZ colorspace is a linear transform of XYZ colorspace is a linear transform of RGB so that all pure wavelengths have RGB so that all pure wavelengths have positive valuespositive values

Page 23: Perception and Measurement of Light, Color, and Appearance

CIE Chromaticity DiagramCIE Chromaticity Diagram

Page 24: Perception and Measurement of Light, Color, and Appearance

Colorspaces for TelevisionColorspaces for Television

• Differences in brightness more Differences in brightness more important than differences in colorimportant than differences in color

• YCYCrrCCbb, YUV, YIQ colorspaces = linear , YUV, YIQ colorspaces = linear

transforms of RGBtransforms of RGB– Lightness: Y=0.299R+0.587G+0.114BLightness: Y=0.299R+0.587G+0.114B

– Other color components typically allocated Other color components typically allocated less bandwidth than Yless bandwidth than Y

Page 25: Perception and Measurement of Light, Color, and Appearance

Perceptually-Uniform ColorspacesPerceptually-Uniform Colorspaces

• Most colorspaces not Most colorspaces not perceptually uniformperceptually uniform

• MacAdam ellipses: MacAdam ellipses: color within each color within each ellipse appears ellipse appears constant (shown constant (shown here 10X size)here 10X size)

Page 26: Perception and Measurement of Light, Color, and Appearance

Perceptually-Uniform ColorspacesPerceptually-Uniform Colorspaces

• u’v’ spaceu’v’ space

• Not perfect, but better than XYZNot perfect, but better than XYZ

ZYX

Yv

ZYX

Xu

315

9'

315

4'

Page 27: Perception and Measurement of Light, Color, and Appearance

L*a*b* Color SpaceL*a*b* Color Space

• Another choice: L*a*b*Another choice: L*a*b*

3/13/1

3/13/1

3/1

200*

500*

16116*

nn

nn

n

Z

Z

Y

Yb

Y

Y

X

Xa

Y

YL

Page 28: Perception and Measurement of Light, Color, and Appearance

L*a*b* Color SpaceL*a*b* Color Space

• Often used for color comparison when Often used for color comparison when “perceptual” differences matter“perceptual” differences matter


Top Related