Zhuo Hui Aswin C Sankaranarayanan Kalyan Sunkavalli Sunil Hadap White balance Image formation albedo light color shading Assumption The scene is Lambertain ID: 563615
Download Presentation The PPT/PDF document "White Balance under Mixed Illumination u..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
White Balance under Mixed Illumination using Flash Photography
Zhuo
HuiAswin C. Sankaranarayanan
Kalyan Sunkavalli Sunil HadapSlide2
White balanceSlide3
Image formation
albedo
light color
shading
Assumption: The scene is
Lambertain
White balance kernel
Slide4
Color chromaticity
=
Chromaticity of albedo
Chromaticity of image intensity
Slide5
Challenges
Multiple illuminants
Single illuminant
Gijsenij
et al.,
Improving color constancy by photometric edge weighting
, PAMI, 2012Slide6
User guidance
Boyadzhiev
et
al
., User-guided white balance for mixed lighting
conditions
, TOG, 2012 Slide7
Flash/no-flash photographySlide8
No-flash image
Flash image
Difference image
Slide9
Image formation
Assumption: flash color is known
Difference imageChromaticity of albedo
Difference image(flash color)Slide10
Chromaticity of no flash
Chromaticity of albedo
Slide11
Chromaticity of no flash
Chromaticity of albedo
Slide12
Shadows
Specularities
MotionsSlide13
White balanced kernels
Flash imageSlide14
Shadow/specularity detection
Relative difference:
Magnitude of difference image
Magnitude of no flash image
Flash energy:
Magnitude of
flash image
No flash energy:
Magnitude of
no-flash image
Shadow
Specularity
large
large
small
small
not too large or too smallSlide15
Camera motion
SIFT + RANSACSlide16Slide17
Scene motion
No flash
Warped no flash
Flash
Forward flow
Backward flowSlide18
Scene motion
No flash
Detected motionsSlide19
Real-world testsSlide20
No flash
s
patially varying lightsSlide21
Results
Assumption:
single dominant lightLambertian scene
Input: single imageO
utput
:
single light color
Angular error:
35.49
◦
Gijsenij
et al.,
Improving color constancy by photometric edge weighting
, PAMI, 2012Slide22
Assumption:
single dominant light
Lambertian sceneknown flash color
Input: flash/no flash
O
utput
:
single light color
Angular error:
14.74
◦
Petschnigg
et al.,
Digital photography with
flash and
no-flash image pairs
, TOG, 2004
ResultsSlide23
Assumption:
Lambertian
sceneInput:
single imageO
utput
:
spatially varying light colors
Angular error:
0.19
◦
Ebner
,
Color constancy using local
color
, ECCV, 2004
ResultsSlide24
Assumption:
two lights
Lambertian scene
Input: single image
known light color
O
utput
:
spatially varying light colors
Angular error:
10.89
◦
Hsu et al.,
Light mixture
estimation for spatially varying white balance
, TOG, 2008
ResultsSlide25
Assumption:
Lambertian
sceneknown flash color
Input: flash/no-flash
O
utput
:
spatially varying light colors
Angular error:
0.88
◦
ResultsSlide26
No-flash
Mixture of outdoor and indoor illuminantsSlide27
Hsu et al.,
Light mixture estimation for spatially varying white balance, TOG, 2008
ResultsAssumption:
two lightsLambertian scene
Input
:
single image
known light color
O
utput
:
spatially varying light colors
Angular error:
20.51
◦Slide28
Results
Assumption:
Lambertian sceneknown flash color
Input: flash/no-flash
O
utput
:
spatially varying light colors
Angular error:
1.18
◦Slide29
No-flashSlide30
Results
Angular error:
1.59◦Slide31
No-flashSlide32
Results
Angular error:
0.12◦Slide33
Shadows/specularitiesSlide34
No flashSlide35
FlashSlide36
Initial white balance resultsSlide37
Detected shadow/specularSlide38
Shadow/specular removal resultsSlide39
Camera and scene motionSlide40
No flashSlide41
FlashSlide42Slide43
Initial white balance resultsSlide44
With rigid alignmentSlide45
With non-rigid compensationSlide46
No flashSlide47
FlashSlide48
White balanced resultsSlide49
LimitationsFails in large violations of the Lambertian
sceneRequires the linear response dataProduces noise estimates for large distance to the cameraSlide50
SummaryWhite balance with flash photography
Automatic, per-pixel and closed form solution