02 fall09 lecture sept18web
TRANSCRIPT
MIT Media LabMIT Media Lab
Camera CultureCamera Culture
Ramesh RaskarRamesh Raskar
MIT Media LabMIT Media Lab
http:// CameraCulture . info/http:// CameraCulture . info/
Computational Camera & Computational Camera & Photography:Photography:
Computational Camera & Computational Camera & Photography:Photography:
Where are the ‘cameras’?Where are the ‘cameras’?
Poll, Sept 18Poll, Sept 18thth 2009 2009
• When will DSCamera disappear? Why?When will DSCamera disappear? Why?
Like Wristwatches ?
Taking NotesTaking Notes
• Use slides I post on the siteUse slides I post on the site
• Write down anecdotes and storiesWrite down anecdotes and stories
• Try to get what is NOT on the slideTry to get what is NOT on the slide
• Summarize questions and answersSummarize questions and answers
• Take photos of demos + doodles on boardTake photos of demos + doodles on board
– Use laptop to take notesUse laptop to take notes– Send before next MondaySend before next Monday
Synthetic LightingSynthetic LightingPaul Haeberli, Jan 1992Paul Haeberli, Jan 1992
HomeworkHomework
• Take multiple photos by changing lighting other parameters. Take multiple photos by changing lighting other parameters. Be creative.Be creative.
• Mix and match color channels to relightMix and match color channels to relight
• Due Sept 25Due Sept 25thth
• Submit on Stellar (via link):Submit on Stellar (via link):– Commented Source codeCommented Source code– Input images and output images PLUS intermediate resultsInput images and output images PLUS intermediate results– CREATE a webpage and send me a linkCREATE a webpage and send me a link
• Ok to use online softwareOk to use online software• Update results on Flickr (group) pageUpdate results on Flickr (group) page
Debevec et al. 2002: ‘Light Stage 3’
Image-Based Actual Re-lightingImage-Based Actual Re-lighting
Film the background in Milan,Film the background in Milan,Measure incoming light,Measure incoming light,
Light the actress in Los AngelesLight the actress in Los Angeles
Matte the backgroundMatte the background
Matched LA and Milan lighting.Matched LA and Milan lighting.
Debevec et al., SIGG2001
Second HomeworkSecond Homework
• Extending Andrew Adam’s Virtual Optical BenchExtending Andrew Adam’s Virtual Optical Bench
Dual photography from diffuse reflections: Dual photography from diffuse reflections: Homework Assignment 2Homework Assignment 2
the camera’s viewSen et al, Siggraph 2005Sen et al, Siggraph 2005
Beyond Visible SpectrumBeyond Visible Spectrum
CedipRedShift
• Brief IntroductionsBrief Introductions• Are you a photographer ?Are you a photographer ?• Do you use camera for vision/image Do you use camera for vision/image processing? Real-time processing?processing? Real-time processing?
• Do you have background in optics/sensors?Do you have background in optics/sensors?
• Name, Dept, Year, Why you are hereName, Dept, Year, Why you are here
• Are you on mailing list? On Stellar?Are you on mailing list? On Stellar?• Did you get email from me?Did you get email from me?
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Dark Bldgs
Reflections on bldgs
Unknown shapes
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
‘Well-lit’ Bldgs
Reflections in bldgs windows
Tree, Street shapes
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, YuBackground is captured from day-time scene using the same fixed camera
Night Image
Day Image
Context Enhanced Image
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Mask is automatically computed from scene contrast
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
But, Simple Pixel Blending Creates Ugly Artifacts
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, YuPixel Blending
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, YuPixel Blending
Our Method:Integration of
blended Gradients
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Rene Magritte, ‘Empire of the Light’
Surrealism
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Time-lapse MosaicsTime-lapse Mosaics
Maggrite StripesMaggrite Stripes
timetime
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
tt
Range Camera DemoRange Camera Demo
http://www.flickr.com/photos/pgoyette/107849943/in/photostream/
Scheimpflug Scheimpflug principleprinciple
PlanPlan
• LensesLenses– Point spread functionPoint spread function
• LightfieldsLightfields– What are they?What are they?– What are the properties?What are the properties?– How to capture?How to capture?– What are the applications?What are the applications?
• FormatFormat– 4 (3) Assignments4 (3) Assignments
• Hands on with optics, Hands on with optics, illumination, sensors, masksillumination, sensors, masks
• Rolling schedule for overlapRolling schedule for overlap• We have cameras, lenses, We have cameras, lenses,
electronics, projectors etcelectronics, projectors etc• Vote on best projectVote on best project
– Mid term examMid term exam• Test conceptsTest concepts
– 1 Final project1 Final project• Should be a Novel and CoolShould be a Novel and Cool• Conference quality paperConference quality paper• Award for best projectAward for best project
– Take 1 class notesTake 1 class notes– Lectures (and guest Lectures (and guest
talks)talks)– In-class + online In-class + online
discussiondiscussion
• If you are a listenerIf you are a listener– Participate in online discussion, dig Participate in online discussion, dig
new recent worknew recent work– Present one short 15 minute idea or Present one short 15 minute idea or
new worknew work
• CreditCredit• Assignments: 40%Assignments: 40%• Project: 30%Project: 30%• Mid-term: 20%Mid-term: 20%• Class participation: 10%Class participation: 10%
• Pre-reqsPre-reqs• Helpful: Linear algebra, image Helpful: Linear algebra, image
processing, think in 3D processing, think in 3D • We will try to keep math to We will try to keep math to
essentials, but complex conceptsessentials, but complex concepts
Assignments:Assignments:You are encouraged to program in Matlab for image analysisYou are encouraged to program in Matlab for image analysisYou may need to use C++/OpenGL/Visual programming for some hardware assignmentsYou may need to use C++/OpenGL/Visual programming for some hardware assignments
Each student is expected to prepare notes for one lectureEach student is expected to prepare notes for one lectureThese notes should be prepared and emailed to the instructor no later than the following These notes should be prepared and emailed to the instructor no later than the following Monday night (midnight EST). Revisions and corrections will be exchanged by email and Monday night (midnight EST). Revisions and corrections will be exchanged by email and after changes the notes will be posted to the website before class the following week.after changes the notes will be posted to the website before class the following week.5 points5 points
Course mailing listCourse mailing list: Please make sure that your emailid is on the course mailing list : Please make sure that your emailid is on the course mailing list Send email to raskar (at) media.mit.eduSend email to raskar (at) media.mit.edu
Please fill in the email/credit/dept sheetPlease fill in the email/credit/dept sheet Office hoursOffice hours::Email is the best way to get in touchEmail is the best way to get in touchRamesh:. raskar (at) media.mit.eduRamesh:. raskar (at) media.mit.eduAnkit: Ankit: ankit (at) media.mit.eduankit (at) media.mit.edu
After class:After class:Muddy Charles PubMuddy Charles Pub(Walker Memorial next to tennis courts)(Walker Memorial next to tennis courts)
2 Sept 18th Modern Optics and Lenses, Ray-matrix operations
3 Sept 25th Virtual Optical Bench, Lightfield Photography, Fourier Optics, Wavefront Coding
4 Oct 2nd Digital Illumination, Hadamard Coded and Multispectral Illumination
5 Oct 9thEmerging Sensors: High speed imaging, 3D range sensors, Femto-second concepts, Front/back
illumination, Diffraction issues
6Oct 16th
Beyond Visible Spectrum: Multispectral imaging and Thermal sensors, Fluorescent imaging, 'Audio camera'
7 Oct 23rdImage Reconstruction Techniques, Deconvolution, Motion and Defocus Deblurring,
Tomography, Heterodyned Photography, Compressive Sensing
8Oct 30th
Cameras for Human Computer Interaction (HCI): 0-D and 1-D sensors, Spatio-temporal coding, Frustrated TIR, Camera-display fusion
9Nov 6th
Useful techniques in Scientific and Medical Imaging: CT-scans, Strobing, Endoscopes, Astronomy and Long range imaging
10Nov 13th
Mid-term Exam, Mobile Photography, Video Blogging, Life logs and Online Photo collections
11Nov 20th Optics and Sensing in Animal Eyes. What can we learn from successful biological vision systems?
12Nov 27th Thanksgiving Holiday (No Class)
13Dec 4th Final Projects
• What are annoyances in photography ?What are annoyances in photography ?
• Why CCD camera behaves retroreflective?Why CCD camera behaves retroreflective?
• Youtube videos on camera tutorial (DoF etc)Youtube videos on camera tutorial (DoF etc)http://www.youtube.com/user/MPTutorhttp://www.youtube.com/user/MPTutor
Anti-Paparazzi FlashAnti-Paparazzi Flash
The anti-paparazzi flash: 1. The celebrity prey. 2. The lurking photographer. 3. The offending camera is detected and then bombed with a beam of light. 4. Voila! A blurry image of nothing much.
• Anti-Paparazzi FlashAnti-Paparazzi Flash
Retroreflective CCD of cellphone camera
Preventing Camera Recording by Designing a Capture-Resistant EnvironmentKhai N. Truong, Shwetak N. Patel, Jay W. Summet, and Gregory D. Abowd.Ubicomp 2005
Auto FocusAuto Focus
• Contrast method compares contrast of images at three depths,Contrast method compares contrast of images at three depths,
if in focus, image will have high contrast, else notif in focus, image will have high contrast, else not
• Phase methods compares two parts of lens at the sensor plane,Phase methods compares two parts of lens at the sensor plane,
if in focus, entire exit pupil sees a uniform color, else notif in focus, entire exit pupil sees a uniform color, else not
• - assumes object has diffuse BRDF- assumes object has diffuse BRDF
Final Project IdeasFinal Project Ideas• User interaction deviceUser interaction device
– Camera basedCamera based
– Illumination basedIllumination based
– Photodetector or line-scan cameraPhotodetector or line-scan camera
• Capture the invisibleCapture the invisible
– Tomography for internalsTomography for internals
– Structured light for 3D scanningStructured light for 3D scanning
– Fluorescence for transparent materialsFluorescence for transparent materials
• Cameras in different EM/other spectrumCameras in different EM/other spectrum
– Wifi, audio, magnetic, haptic, capacitiveWifi, audio, magnetic, haptic, capacitive
– Visible Thermal IR segmentationVisible Thermal IR segmentation
– Thermal IR (emotion detection, motion Thermal IR (emotion detection, motion detector)detector)
– Multispectral camera, discriminating (camel-Multispectral camera, discriminating (camel-sand)sand)
• IlluminationIllumination
– Multi-flash with lighfieldMulti-flash with lighfield
– Schielren photographySchielren photography
– Strobing and Colored strobingStrobing and Colored strobing
• External non-imaging sensorExternal non-imaging sensor– Camera with gyro movement sensors, Camera with gyro movement sensors,
find identity of userfind identity of user– Cameras with GPS and online geo-tagged Cameras with GPS and online geo-tagged
photo collectionsphoto collections– Interaction between two cameras (with Interaction between two cameras (with
lasers on-board)lasers on-board)• OpticsOptics
– LightfieldLightfield– Coded apertureCoded aperture– Bio-inspired visionBio-inspired vision
• TimeTime– Time-lapse photosTime-lapse photos– Motion blurMotion blur
Kitchen Sink: Volumetric Scattering
Direct Global
Volumetric Scattering:
Chandrasekar 50, Ishimaru 78
“Origami Lens”: Thin Folded Optics (2007)
“Ultrathin Cameras Using Annular Folded Optics, “E. J. Tremblay, R. A. Stack, R. L. Morrison, J. E. FordApplied Optics, 2007 - OSA
Slides by Shree Nayar
Origami Lens
ConventionalLens
Origami Lens
Slides by Shree Nayar
Fernald, Science [Sept 2006]
Shadow Refractive
Reflective
Tools
for
Visual Computin
g
Photonic Crystals• ‘Routers’ for photons instead of electrons
• Photonic Crystal– Nanostructure material with ordered array of holes– A lattice of high-RI material embedded within a lower RI – High index contrast– 2D or 3D periodic structure
• Photonic band gap– Highly periodic structures that blocks certain wavelengths– (creates a ‘gap’ or notch in wavelength)
• Applications– ‘Semiconductors for light’: mimics silicon band gap for electrons– Highly selective/rejecting narrow wavelength filters (Bayer Mosaic?)– Light efficient LEDs– Optical fibers with extreme bandwidth (wavelength multiplexing)– Hype: future terahertz CPUs via optical communication on chip
Schlieren Photography
• Image of small index of refraction gradients in a gas• Invisible to human eye (subtle mirage effect)
Knife edge blocks half the light unless
distorted beam focuses imperfectly
Collimated Light
Camera
http://www.mne.psu.edu/psgdl/FSSPhotoalbum/index1.htm
Sample Final ProjectsSample Final Projects
• Schlieren Photography Schlieren Photography – (Best project award + Prize in 2008)(Best project award + Prize in 2008)
• Camera array for Particle Image VelocimetryCamera array for Particle Image Velocimetry
• BiDirectional ScreenBiDirectional Screen
• Looking Around a Corner (theory)Looking Around a Corner (theory)
• Tomography machineTomography machine
• ....
• ....
Computational IlluminationComputational Illumination
• Dual PhotographyDual Photography
• Direct-global SeparationDirect-global Separation
• Multi-flash CameraMulti-flash Camera
Ramesh Raskar, Computational Illumination
Computational Illumination
Computational Computational PhotographyPhotography
Illumination
Novel CamerasGeneralized
Sensor
Generalized Optics
Processing
4D Light Field
Computational IlluminationComputational Illumination
Novel CamerasGeneralized
Sensor
Generalized Optics
Processing
Generalized Optics
Light Sources
Modulators
4D Light Field
Programmable 4D Illumination field + time + wavelength
Programmable 4D Illumination field + time + wavelength
Novel Illumination
Edgerton 1930’sEdgerton 1930’s
Not Special Cameras but Special Lighting
Edgerton 1930’sEdgerton 1930’s
Multi-flash Sequential Photography
Stroboscope(Electronic Flash)
Shutter Open
Flash Time
‘Smarter’ Lighting Equipment
What Parameters Can We What Parameters Can We Change ?Change ?
Computational Illumination:Computational Illumination:Programmable 4D Illumination Field + Time + WavelengthProgrammable 4D Illumination Field + Time + Wavelength
• Presence or Absence, Duration, BrightnessPresence or Absence, Duration, Brightness– Flash/No-flashFlash/No-flash
• Light positionLight position– Relighting: Programmable domeRelighting: Programmable dome– Shape enhancement: Multi-flash for depth edgesShape enhancement: Multi-flash for depth edges
• Light color/wavelengthLight color/wavelength• Spatial ModulationSpatial Modulation
– Synthetic Aperture IlluminationSynthetic Aperture Illumination
• Temporal ModulationTemporal Modulation– TV remote, Motion Tracking, Sony ID-cam, RFIGTV remote, Motion Tracking, Sony ID-cam, RFIG
• Exploiting (uncontrolled) natural lighting conditionExploiting (uncontrolled) natural lighting condition– Day/Night Fusion, Time Lapse, GlareDay/Night Fusion, Time Lapse, Glare
Multi-flash Camera for Detecting Depth Edges
Ramesh Raskar, Karhan Tan, Rogerio Feris, Jingyi Yu, Matthew Turk
Mitsubishi Electric Research Labs (MERL), Cambridge, MAU of California at Santa Barbara
U of North Carolina at Chapel Hill
Non-photorealistic Camera: Non-photorealistic Camera: Depth Edge Detection Depth Edge Detection andand Stylized Stylized
Rendering Rendering usingusing Multi-Flash ImagingMulti-Flash Imaging
Car Manuals
What are the problems with ‘real’ photo in conveying information ?
Why do we hire artists to draw what can be photographed ?
Shadows
Clutter
Many Colors
Highlight Shape Edges
Mark moving parts
Basic colors
Shadows
Clutter
Many Colors
Highlight Edges
Mark moving parts
Basic colors
A New ProblemA New Problem
GesturesGestures
Input Photo Canny Edges Depth Edges
Depth Edges with MultiFlashDepth Edges with MultiFlashRaskar, Tan, Feris, Jingyi Yu, Turk – ACM SIGGRAPH 2004
Depth Discontinuities
Internal and externalShape boundaries, Occluding contour, Silhouettes
Depth Edges
Our MethodCanny
Canny Intensity Edge Detection
Our Method
Photo Result
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Imaging Geometry
Shadow lies along epipolar ray
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Shadow lies along epipolar ray,
Epipole and Shadow are on opposite sides of the edge
Imaging Geometry
m
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Shadow lies along epipolar ray,
Shadow and epipole are on opposite sides of the edge
Imaging Geometry
m
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Depth Edge Camera
Light epipolar rays are horizontal or vertical
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Normalized
Left / Max
Right / Max
Left Flash
Right Flash
Input U{depth edges}
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Normalized
Left / Max
Right / Max
Left Flash
Right Flash
Input U{depth edges}
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Normalized
Left / Max
Right / Max
Left Flash
Right Flash
Input U{depth edges}
Negative transition along epipolar ray is depth edge
Plot
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
Normalized
Left / Max
Right / Max
Left Flash
Right Flash
Input
Negative transition along epipolar ray is depth edge
Plot U{depth edges}
Mitsubishi Electric Research Labs Raskar, Tan, Feris, Yu, TurkMultiFlash NPR Camera
% Max composite maximg = max( left, right, top, bottom);
% Normalize by computing ratio imagesr1 = left./ maximg; r2 = top ./ maximg;r3 = right ./ maximg; r4 = bottom ./ maximg;
% Compute confidence mapv = fspecial( 'sobel' ); h = v';d1 = imfilter( r1, v ); d3 = imfilter( r3, v ); % vertical sobeld2 = imfilter( r2, h ); d4 = imfilter( r4, h ); % horizontal sobel
%Keep only negative transitions silhouette1 = d1 .* (d1>0); silhouette2 = abs( d2 .* (d2<0) );silhouette3 = abs( d3 .* (d3<0) );silhouette4 = d4 .* (d4>0);
%Pick max confidence in eachconfidence = max(silhouette1, silhouette2, silhouette3, silhouette4);imwrite( confidence, 'confidence.bmp');
No magicparameters
!
Depth Edges
Left Top Right Bottom
Depth EdgesCanny Edges
GesturesGestures
Input Photo Canny Edges Depth Edges
Flash MattingFlash Matting
Flash Matting, Jian Sun, Yin Li, Sing Bing Kang, and Heung-Yeung Shum, Siggraph 2006
Multi-light Image CollectionMulti-light Image Collection[Fattal, Agrawala, Rusinkiewicz] Sig’2007
Input Photos ShadowFree,Enhanced
surface detail,
but Flat look
Some Shadows for depth
but Lost visibility
Multiscale decomposition using Bilateral Filter,Combine detail at each scale across all the input images.
Fuse maximum gradient from each photo, Reconstruct from 2D integrationall the input images.
Enhanced shadows
Computational Illumination:Computational Illumination:Programmable 4D Illumination Field + Time + WavelengthProgrammable 4D Illumination Field + Time + Wavelength
• Presence or Absence, Duration, BrightnessPresence or Absence, Duration, Brightness– Flash/No-flash (matting for foreground/background)Flash/No-flash (matting for foreground/background)
• Light positionLight position– Relighting: Programmable domeRelighting: Programmable dome– Shape enhancement: Multi-flash for depth edgesShape enhancement: Multi-flash for depth edges
• Light color/wavelengthLight color/wavelength• Spatial ModulationSpatial Modulation
– Dual Photography, Direct/Global Separation, Synthetic Dual Photography, Direct/Global Separation, Synthetic Aperture IlluminationAperture Illumination
• Temporal ModulationTemporal Modulation– TV remote, Motion Tracking, Sony ID-cam, RFIGTV remote, Motion Tracking, Sony ID-cam, RFIG
• Exploiting (uncontrolled) natural lighting conditionExploiting (uncontrolled) natural lighting condition– Day/Night Fusion, Time Lapse, GlareDay/Night Fusion, Time Lapse, Glare
Los Angeles, CA August 2, 2005 Pradeep Sen
Dual Photography
Pradeep Sen, Billy Chen, Gaurav Garg, Steve MarschnerMark Horowitz, Marc Levoy, Hendrik Lensch
Stanford University Cornell University
*
*
August 2, 2005
Los Angeles, CA
Los Angeles, CA August 2, 2005 Pradeep Sen
The card experiment
book
camera
card
projector
primal
Los Angeles, CA August 2, 2005 Pradeep Sen
The card experiment
primal
dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Overview of dual photography
standard photographfrom camera
dual photographfrom projector
Los Angeles, CA August 2, 2005 Pradeep Sen
Outline
1. Introduction to dual photography
2. Application to scene relighting
3. Accelerating acquisition
4. Conclusions
Los Angeles, CA August 2, 2005 Pradeep Sen
Helmholtz reciprocity
light
eye
eye
light I
I
I
I
primaldual
scene
projector
photosensor
primal
Los Angeles, CA August 2, 2005 Pradeep Sen
Helmholtz reciprocity
projector
photosensor
projector
photosensor
scene
light
cameraprimaldual
Los Angeles, CA August 2, 2005 Pradeep Sen
Forming a dual image
projector
photosensor C0
C1
C2
C3
C4
C5
C6
C7
light
camera
scene
primaldual
Los Angeles, CA August 2, 2005 Pradeep Sen
Forming a dual image
scene
C0
C1
C2
C3
C4
C5
C6
C7
light
cameradual
Los Angeles, CA August 2, 2005 Pradeep Sen
Physical demonstration
Projector was scanned across a scene while a photosensor measured the outgoing light
photosensor
resulting dual image
Los Angeles, CA August 2, 2005 Pradeep Sen
Related imaging methods
Example of a “flying-spot” camera built at the dawn of TV (Baird 1926)
Scanning electron microscope
Velcro® at 35x magnification,Museum of Science, Boston
Los Angeles, CA August 2, 2005 Pradeep Sen
Dual photography for relighting
p
q n
m
projector cameradual camera dual projector
4D
projectorphotosensor
primal
scene
dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Mathematical notation
scene
p
q n
m
projector camera
P
pq x 1
C
mn x 1
T
mn x pq
primal
Los Angeles, CA August 2, 2005 Pradeep Sen
Mathematical notation
P
pq x 1
C
mn x 1
T
mn x pq
=
primal
Los Angeles, CA August 2, 2005 Pradeep Sen
=
pq x 1mn x 1
T
mn x pq
Mathematical notation
100000
Los Angeles, CA August 2, 2005 Pradeep Sen
=
pq x 1mn x 1
T
mn x pq
Mathematical notation
010000
Los Angeles, CA August 2, 2005 Pradeep Sen
=
pq x 1mn x 1
T
mn x pq
Mathematical notation
001000
Los Angeles, CA August 2, 2005 Pradeep Sen
= T
mn x pq
Mathematical notation
P
pq x 1
C
mn x 1little interreflection → sparse matrix
many interreflections → dense matrix
Los Angeles, CA August 2, 2005 Pradeep Sen
?
Mathematical notation
T= P
pq x 1
C
mn x 1
mn x pq
primal space
dual space =
pq x mn
C
mn x 1
P
pq x 1
j
i
i
j
Tij
T
Tji
T = TT
T = Tji ij
Los Angeles, CA August 2, 2005 Pradeep Sen
Definition of dual photography
= T
mn x pq
primal space
dual space =
pq x mn
TT
T
mn x pq
mn x 1
C
pq x 1
P
pq x 1
P
mn x 1
C
Los Angeles, CA August 2, 2005 Pradeep Sen
Sample results
primal dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Sample results
primal dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Scene relighting
Knowing the pixel-to-pixel transport between the projector and the camera allows us to relight the scene with an arbitrary 2D pattern
primal dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Photosensor experiment
= T
mn x pq
pq x 1
P
mn x 1
C C T
1 x pq
primal space
dual space
T
1 x pq
=
pq x 1
P
pq x 1
TT
C
Los Angeles, CA August 2, 2005 Pradeep Sen
2D relighting videos
Relighting book scene with animated patterns
Relighting box with animated pattern
Los Angeles, CA August 2, 2005 Pradeep Sen
Relighting with 4D incident light fields
2D 4D
6D Transport
Los Angeles, CA August 2, 2005 Pradeep Sen
From Masselus et al. SIGGRAPH ‘03
Relighting with 4D incident light fields
Los Angeles, CA August 2, 2005 Pradeep Sen
Relighting with 4D incident light fields
2D 4D
6D Transport
Los Angeles, CA August 2, 2005 Pradeep Sen
Relighting with 4D incident light fields
2D 4D
6D Transport
Los Angeles, CA August 2, 2005 Pradeep Sen
Acquisition of transport from multiple projectors cannot be parallelized
Acquisition of transport using multiple cameras can!
Advantages of our dual framework
Los Angeles, CA August 2, 2005 Pradeep Sen
“Multiple” cameras with mirror array
Los Angeles, CA August 2, 2005 Pradeep Sen
Conclusions
Dual photography is a novel imaging technique that allows us to interchange the camera and the projector
Dual photography can accelerate acquisition of the 6D transport for relighting with 4D incident light fields
We developed an algorithm to accelerate the acquisition of the transport matrix for dual photography
Los Angeles, CA August 2, 2005 Pradeep Sen
The card experiment
book
camera
card
projector
primal
Los Angeles, CA August 2, 2005 Pradeep Sen
The card experiment
primal
dual
Los Angeles, CA August 2, 2005 Pradeep Sen
Photosensor experiment
= T
mn x pq
pq x 1
P
mn x 1
C C T
1 x pq
primal space
dual space
T
1 x pq
=
pq x 1
P
pq x 1
TT
C
Computational Illumination:Computational Illumination:Programmable 4D Illumination Field + Time + WavelengthProgrammable 4D Illumination Field + Time + Wavelength
• Presence or Absence, Duration, BrightnessPresence or Absence, Duration, Brightness– Flash/No-flash (matting for foreground/background)Flash/No-flash (matting for foreground/background)
• Light positionLight position– Relighting: Programmable domeRelighting: Programmable dome– Shape enhancement: Multi-flash for depth edgesShape enhancement: Multi-flash for depth edges
• Light color/wavelengthLight color/wavelength• Spatial ModulationSpatial Modulation
– Dual Photography, Direct/Global Separation, Synthetic Dual Photography, Direct/Global Separation, Synthetic Aperture IlluminationAperture Illumination
• Temporal ModulationTemporal Modulation– TV remote, Motion Tracking, Sony ID-cam, RFIGTV remote, Motion Tracking, Sony ID-cam, RFIG
• Exploiting (uncontrolled) natural lighting conditionExploiting (uncontrolled) natural lighting condition– Day/Night Fusion, Time Lapse, GlareDay/Night Fusion, Time Lapse, Glare
Visual Chatter in the Real World
Shree K. Nayar
Computer ScienceColumbia University
With: Guru Krishnan, Michael Grossberg, Ramesh Raskar
Eurographics Rendering SymposiumJune 2006, Nicosia, Cyprus
Support: ONR
source
surface
P
Direct and Global Illumination
A
A : Direct
B
B : Interrelection
C
C : Subsurface
D
participating medium
D : Volumetric translucent surface
E
E : Diffusion
camera
Direct Global
Shower Curtain: Diffuser
],[],[],[ icLicLicL gd
direct globalradiance
Direct and Global Components: Interreflections
surface
i
camera
source
P
g jiLjiAicL ],[],[],[
j
BRDF and geometry
High Frequency Illumination Pattern
surface
camera
source
fraction of activated source elements
],[],[],[ icLicLicL gd +
i
High Frequency Illumination Pattern
surface
fraction of activated source elements
camera
source
],[],[],[ icLicLicL gd + - ],[],[ icLicL g )1(
i
:2
1 min2LLg
Separation from Two Images
direct global
,minmax LLLd
Other Global Effects: Volumetric Scattering
surface
camera
source
participating medium
i
j
Diffuse Interreflections
SpecularInterreflections
Volumetric Scattering Subsurface
Scattering
Diffusion
Scene
Direct Global
F G C A D B E
A: Diffuse Interreflection (Board)
B: Specular Interreflection (Nut)
C: Subsurface Scattering (Marble)
D: Subsurface Scattering (Wax)
E: Translucency (Frosted Glass)
F: Volumetric Scattering (Dilute Milk)
G: Shadow (Fruit on Board)
Verification
V-Grooves: Diffuse Interreflections
Direct Global
concave convex
Psychophysics:
Gilchrist 79, Bloj et al. 04
Mirror Ball: Failure Case
Direct Global
Real World Examples:
Can You Guess the Images?
Eggs: Diffuse Interreflections
Direct Global
Stick
Building Corner
Shadow
minLLg
direct global
,minmax LLLd
3D from Shadows:
Bouguet and Perona 99
Direct Global
Building Corner
Shower Curtain: Diffuser
Shadow
Mesh
minLLg
direct global
,minmax LLLd
Direct Global
Shower Curtain: Diffuser
Kitchen Sink: Volumetric Scattering
Direct Global
Volumetric Scattering:
Chandrasekar 50, Ishimaru 78
Novel Image
Real or Fake ?
Direct Global
R
F
RF
R
F
Hand
Direct Global
Skin: Hanrahan and Krueger 93,
Uchida 96, Haro 01, Jensen et al. 01,
Igarashi et al. 05, Weyrich et al. 05
Hands
Afric. Amer.Female
ChineseMale
SpanishMale
Direct Global
Afric. Amer.Female
ChineseMale
SpanishMale
Afric. Amer.Female
ChineseMale
SpanishMale
Pink Carnation
GlobalDirect
Spectral Bleeding: Funt et al. 91
Summary
• Fast and Simple Separation Method
• Wide Variety of Global Effects
• No Prior Knowledge of Material Properties
• Implications:
• Generation of Novel Images
• Enhance Computer Vision Methods
• Insights into Properties of Materials
+=
+=
+=
+=
+=
+=
+
+
=
=
+=
+=
www.cs.columbia.edu/CAVE
Computational Illumination:Computational Illumination:Programmable 4D Illumination Field + Time + WavelengthProgrammable 4D Illumination Field + Time + Wavelength
• Presence or Absence, Duration, BrightnessPresence or Absence, Duration, Brightness– Flash/No-flash (matting for foreground/background)Flash/No-flash (matting for foreground/background)
• Light positionLight position– Relighting: Programmable domeRelighting: Programmable dome– Shape enhancement: Multi-flash for depth edgesShape enhancement: Multi-flash for depth edges
• Light color/wavelengthLight color/wavelength• Spatial ModulationSpatial Modulation
– Dual Photography, Direct/Global Separation, Synthetic Dual Photography, Direct/Global Separation, Synthetic Aperture IlluminationAperture Illumination
• Temporal ModulationTemporal Modulation– TV remote, Motion Tracking, Sony ID-cam, RFIGTV remote, Motion Tracking, Sony ID-cam, RFIG
• Exploiting (uncontrolled) natural lighting conditionExploiting (uncontrolled) natural lighting condition– Day/Night Fusion, Time Lapse, GlareDay/Night Fusion, Time Lapse, Glare
Day of the year
Time of day
The Archive of Many Outdoor Scenes (AMOS)
Images from ~1000 static webcams, every 30 minutes since March 2006.
Variations over a year and over a day
Jacobs, Roman, and Robert Pless, WUSTL CVPR 2007,
Analysing Time Lapse Images
• PCA– Linear Variations due to lighting and seasonal variation
• Decompose (by time scale)– Hour: haze and cloud for depth.
– Day: changing lighting directions for surface orientation
– Year: effects of changing seasons highlight vegetation
• Applications:– Scene segmentation.– Global Webcam localization. Correlate timelapse
video over a month from unknown camera with:• sunrise + sunset (localization accuracy ~ 50 miles)
• Known nearby cameras (~25 miles)
• Satellite image (~15 miles)
Mean image + 3 components from time lapse of downtown st. louis over the course of 2 hours
2 Hour time Lapse in St Louis:Depth from co-varying regions
Surface Orientation
False Color PCA images
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Image Fusion for Context Enhancementand Video Surrealism
Adrian IlieAdrian IlieRamesh RaskarRamesh Raskar Jingyi YuJingyi Yu
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
Dark Bldgs
Reflections on bldgs
Unknown shapes
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, Yu
‘Well-lit’ Bldgs
Reflections in bldgs windows
Tree, Street shapes
Mitsubishi Electric Research Laboratories Image Fusion for Context Enhancement Raskar, Ilie, YuBackground is captured from day-time scene using the same fixed camera
Night Image
Day Image
Context Enhanced Image
http://web.media.mit.edu/~raskar/NPAR04/
Ramesh Raskar, CompPhoto
Factored Time Lapse Video
Factor into shadow, illumination, and reflectance. Relight, recover surface normals, reflectance editing.
[Sunkavalli, Matusik, Pfister, Rusinkiewicz], Sig’07
Ramesh Raskar, CompPhoto
Computational Illumination:Computational Illumination:Programmable 4D Illumination Field + Time + WavelengthProgrammable 4D Illumination Field + Time + Wavelength
• Presence or Absence, Duration, BrightnessPresence or Absence, Duration, Brightness– Flash/No-flash (matting for foreground/background)Flash/No-flash (matting for foreground/background)
• Light positionLight position– Relighting: Programmable domeRelighting: Programmable dome– Shape enhancement: Multi-flash for depth edgesShape enhancement: Multi-flash for depth edges
• Light color/wavelengthLight color/wavelength• Spatial ModulationSpatial Modulation
– Dual Photography, Direct/Global Separation, Synthetic Dual Photography, Direct/Global Separation, Synthetic Aperture IlluminationAperture Illumination
• Temporal ModulationTemporal Modulation– TV remote, Motion Tracking, Sony ID-cam, RFIGTV remote, Motion Tracking, Sony ID-cam, RFIG
• Exploiting (uncontrolled) natural lighting conditionExploiting (uncontrolled) natural lighting condition– Day/Night Fusion, Time Lapse, GlareDay/Night Fusion, Time Lapse, Glare