11/01/06Jean-François Lalonde Natural Color Statistics p. 1 Natural color statistics Jean-François Lalonde Misc-read, November 1 st 2006.

Slides:



Advertisements
Similar presentations
A Common Framework for Ambient Illumination in the Dichromatic Reflectance Model Color and Reflectance in Imaging and Computer Vision Workshop 2009 October.
Advertisements

Bayesian Belief Propagation
Color Image Understanding Sharon Alpert & Denis Simakov.
Color Harmonization - ACM SIGGRAPH 2006 Speaker :李沃若.
 Image Characteristics  Image Digitization Spatial domain Intensity domain 1.
Image Processing IB Paper 8 – Part A Ognjen Arandjelović Ognjen Arandjelović
Color spaces CIE - RGB space. HSV - space. CIE - XYZ space.
Color2Gray: Salience-Preserving Color Removal
Color2Gray: Salience-Preserving Color Removal Amy Gooch Sven Olsen Jack Tumblin Bruce Gooch Northwestern University.
Color Image Processing
Physics-based Illuminant Color Estimation as an Image Semantics Clue Christian Riess Elli Angelopoulou Pattern Recognition Lab (Computer Science 5) University.
University of British Columbia CPSC 314 Computer Graphics Jan-Apr 2007 Tamara Munzner Vision/Color II, Virtual.
School of Computing Science Simon Fraser University
School of Computer Science Simon Fraser University November 2009 Sharpening from Shadows: Sensor Transforms for Removing Shadows using a Single Image Mark.
Sensor Transforms to Improve Metamerism-Based Watermarking Mark S. Drew School of Computing Science Simon Fraser University Raja Bala Xerox.
Retinex Image Enhancement Techniques --- Algorithm, Application and Advantages Prepared by: Zhixi Bian and Yan Zhang.
1 Color Segmentation: Color Spaces and Illumination Mohan Sridharan University of Birmingham
Apparent Greyscale: A Simple and Fast Conversion to Perceptually Accurate Images and Video Kaleigh SmithPierre-Edouard Landes Joelle Thollot Karol Myszkowski.
Perceptual Evaluation of Colour Gamut Mapping Algorithms Fabienne Dugay The Norwegian Color Research Laboratory Faculty of Computer Science and Media Technology.
Color Image Understanding Sharon Alpert & Denis Simakov.
Capturing Light… in man and machine : Computational Photography Alexei Efros, CMU, Fall 2006 Some figures from Steve Seitz, Steve Palmer, Paul Debevec,
What is color for?.
Color, lightness & brightness Lavanya Sharan February 7, 2011.
Capturing Light… in man and machine : Computational Photography Alexei Efros, CMU, Fall 2010.
CS292 Computational Vision and Language Visual Features - Colour and Texture.
Introduction of the intrinsic image. Intrinsic Images The method of Finlayson & Hordley ( 2001 ) Two assumptions 1. the camera ’ s sensors are sufficiently.
Colour Image Compression by Grey to Colour Conversion Mark S. Drew 1, Graham D. Finlayson 2, and Abhilash Jindal 3 1 Simon Fraser University 2 University.
CS559-Computer Graphics Copyright Stephen Chenney Color Recap The physical description of color is as a spectrum: the intensity of light at each wavelength.
Power Minimization for LED-backlit TFT-LCDs Wei-Chung Cheng July 26, 2006 PODLAB – Perception Oriented Design Lab Department of Photonics and Display.
9/14/04© University of Wisconsin, CS559 Spring 2004 Last Time Intensity perception – the importance of ratios Dynamic Range – what it means and some of.
Face Recognition and Feature Subspaces
Face Recognition and Feature Subspaces
1 © 2010 Cengage Learning Engineering. All Rights Reserved. 1 Introduction to Digital Image Processing with MATLAB ® Asia Edition McAndrew ‧ Wang ‧ Tseng.
Computer Science 631 Lecture 7: Colorspace, local operations
Color Principles for Computer Graphics Donald House 9/17/09 Artist’s slides by Lynette House.
November 2012 The Role of Bright Pixels in Illumination Estimation Hamid Reza Vaezi Joze Mark S. Drew Graham D. Finlayson Petra Aurora Troncoso Rey School.
Topic 5 - Imaging Mapping - II DIGITAL IMAGE PROCESSING Course 3624 Department of Physics and Astronomy Professor Bob Warwick.
Color Theory ‣ What is color? ‣ How do we perceive it? ‣ How do we describe and match colors? ‣ Color spaces.
03/05/03© 2003 University of Wisconsin Last Time Tone Reproduction If you don’t use perceptual info, some people call it contrast reduction.
How natural scenes might shape neural machinery for computing shape from texture? Qiaochu Li (Blaine) Advisor: Tai Sing Lee.
Computer Vision Lab. SNU Young Ki Baik Nonlinear Dimensionality Reduction Approach (ISOMAP, LLE)
Why is computer vision difficult?
CIE Budapest 2009 | FG Lichttechnik | Peter Bodrogi et al. Re-defining the colour rendering index Peter Bodrogi Stefan Brückner Tran Quoc Khanh.
TVCG 2013 Sungkil Lee, Mike Sips, and Hans-Peter Seidel.
Modeling the Shape of a Scene: Seeing the trees as a forest Scene Understanding Seminar
Color Processing : Rendering and Image Processing Alexei Efros …with most figures shamelessly stolen from Forsyth & Ponce and Gonzalez & Woods.
Sensory Information Processing
CS654: Digital Image Analysis Lecture 29: Color Image Processing.
A NOVEL METHOD FOR COLOR FACE RECOGNITION USING KNN CLASSIFIER
Color Principles for Computer Graphics Donald House 9/17/09 Artist’s slides by Lynette House.
Color and Brightness Constancy Jim Rehg CS 4495/7495 Computer Vision Lecture 25 & 26 Wed Oct 18, 2002.
ECE 638: Principles of Digital Color Imaging Systems Lecture 4: Chromaticity Diagram.
David Luebke 1 2/5/2016 Color CS 445/645 Introduction to Computer Graphics David Luebke, Spring 2003.
ECE 638: Principles of Digital Color Imaging Systems Lecture 12: Characterization of Illuminants and Nonlinear Response of Human Visual System.
Learning and Removing Cast Shadows through a Multidistribution Approach Nicolas Martel-Brisson, Andre Zaccarin IEEE TRANSACTIONS ON PATTERN ANALYSIS AND.
1 of 32 Computer Graphics Color. 2 of 32 Basics Of Color elements of color:
Color Models Light property Color models.
Capturing Light… in man and machine
Capturing Light… in man and machine
Chapter 6: Color Image Processing
ECE 638: Principles of Digital Color Imaging Systems
Computer Vision Lecture 4: Color
Outline Color perception Introduction Theories of color perception
Introduction to Perception and Color
Capturing Light… in man and machine
Slides taken from Scott Schaefer
Digital Image Processing
Color Model By : Mustafa Salam.
COLOR CONSTANCY IN THE COMPRESSED DOMAIN
Color Theory What is color? How do we perceive it?
Presentation transcript:

11/01/06Jean-François Lalonde Natural Color Statistics p. 1 Natural color statistics Jean-François Lalonde Misc-read, November 1 st 2006

11/01/06Jean-François Lalonde Natural Color Statistics p. 2 Image formation & capture BRDF Irradiance Sensor's response Radiance

11/01/06Jean-François Lalonde Natural Color Statistics p. 3 Color spaces RGB  Used in displays HSV  Separates luminance from chroma from “purity” CIE L*a*b*  Separates luminance from chroma  Close to perceptual uniformity R G B

11/01/06Jean-François Lalonde Natural Color Statistics p. 4 Big Picture Is color really important?

11/01/06Jean-François Lalonde Natural Color Statistics p. 5 Is color important? Colors contribute to recognition when they are diagnostic of a scene category [Oliva & Synchs, 2000] Diagnostic Colors Mediate Scene Recognition DiagnosticNon-diagnostic

11/01/06Jean-François Lalonde Natural Color Statistics p. 6 Experiment 3 – phase 1

11/01/06Jean-François Lalonde Natural Color Statistics p. 7 Experiment 3 – phase 1

11/01/06Jean-François Lalonde Natural Color Statistics p. 8 Experiment #2 – phase 2

11/01/06Jean-François Lalonde Natural Color Statistics p. 9 Experiment #2 – phase 2

11/01/06Jean-François Lalonde Natural Color Statistics p. 10 Color is important! 3 Experiments  Colors contribute to recognition when they are diagnostic of a scene category  Faster verification of category membership of scenes when properly colored  Addition of colors to coarse luminance blobs enhance categorization [Olivia & Torralba, 2006] Building the Gist of a Scene: The Role of Global Image Features in Recognition

11/01/06Jean-François Lalonde Natural Color Statistics p. 11 Big Picture – (bis) Is color really important? Can natural color be compactly represented? [Oliva & Synchs, 2000] [Oliva & Torralba, 2006] YES! important for scene recognition

11/01/06Jean-François Lalonde Natural Color Statistics p channels? Human cones: L,M,S color receptors

11/01/06Jean-François Lalonde Natural Color Statistics p channels Low dimensionality of natural reflectances  PCA on spectral data over visible range  98% of energy can be represented using 3 components [Chiao & Cronin, 2000] Coral reefForest

11/01/06Jean-François Lalonde Natural Color Statistics p. 14 Big Picture – revisited Can natural color be compactly represented? [Chiao & Cronin, 2000] How?? [Oliva & Synchs, 2000] [Oliva & Torralba, 2006] YES! important for scene recognition 3 channels account for 98% variability Is color really important?

11/01/06Jean-François Lalonde Natural Color Statistics p. 15 Mapping to know illuminant Color constancy: reduction of the effect of the scene illumination  Recover color under known illuminant Not true object reflectance!  Estimate mapping from unknown  known Linear model  C known = A  C unknown

11/01/06Jean-François Lalonde Natural Color Statistics p. 16 Color constancy setup Known (canonical) illuminant Unknown scene Unknown illuminant

11/01/06Jean-François Lalonde Natural Color Statistics p. 17 Classic paper: Gamut mapping Gamut: convex set [Forsyth, 1990], [Barnard, 1998] Known illuminant Canonical gamut Unknown illuminant a b c B A C aAaA aCaC aBaB Unknown gamut Canonical gamut

11/01/06Jean-François Lalonde Natural Color Statistics p. 18 Gamut mapping: transformations cBcB cAcA cCcC bAbA bCbC bBbB aBaB aCaC aAaA [Forsyth, 1990], [Barnard, 1998], [Finlayson, 1995] Possible to do it in chromaticity space (2-D)

11/01/06Jean-François Lalonde Natural Color Statistics p. 19 G.D. Finlayson ~170 papers on color constancy  Somewhat incremental Color by Correlation  Colors in an image provide information about the illuminant [Finlayson et al., 2001], [Schaefer et al., 2005]

11/01/06Jean-François Lalonde Natural Color Statistics p. 20 Color by Correlation  RGB  chromaticity (no luminance)  Characterize illuminants  Matrix M(c, l) = P(chromaticity c | light l ) 3. Quantize input image 4. Compute correlation 5. Select best illuminant  Max correlation [Finlayson et al., 2001], [Schaefer et al., 2005] Gamut mapping performs similar to using P(c|l)  {0,1}

11/01/06Jean-François Lalonde Natural Color Statistics p. 21 Color constancy Major problem: require calibrated illuminants! Alternative: color flows  How color commonly change together under natural illuminant variation Allow non-linear transformations Data-driven [Miller & Tieu, 2001]

11/01/06Jean-François Lalonde Natural Color Statistics p. 22 Color flows Kernel density estimator Partially-observed color flow Full color flow Slice of RGB cube

11/01/06Jean-François Lalonde Natural Color Statistics p. 23 Color eigenflows Subsample RGB cube Apply PCA, keep first k eigenflows First 3 eigenflows: Original image [Miller & Tieu, 2001]

11/01/06Jean-François Lalonde Natural Color Statistics p. 24 Big Picture – yet again Can natural color be compactly represented? [Chiao & Cronin, 2000] How?? [Barnard, 1998] [Forsyth, 1990] [Finlayson, 1995] [Finlayson et al., 2001] [Miller & Tieu, 2001] [Schaefer et al., 2005] [Oliva & Synchs, 2000] [Oliva & Torralba, 2006] YES! important for scene recognition 3 channels account for 98% variability Mapping to illuminant (color constancy) Is color really important?

11/01/06Jean-François Lalonde Natural Color Statistics p. 25 Color harmony Harmonic colors  Aesthetically pleasing in terms of human visual perception Matsuda's regions [Matsuda, 1995], [Tokumaru et al., 2002],[Cohen-Or et al., 2006]

11/01/06Jean-François Lalonde Natural Color Statistics p. 26 Color harmonization Find best-fitting template Squash all colors inside template (hue only) [Cohen-Or et al., 2006]

11/01/06Jean-François Lalonde Natural Color Statistics p. 27 Color harmonization [Cohen-Or et al., 2006] Before Harmonizing background with foreground Does not change natural images!

11/01/06Jean-François Lalonde Natural Color Statistics p. 28 Color harmonization

11/01/06Jean-François Lalonde Natural Color Statistics p. 29 Big Picture – last time I promise Can natural color be compactly represented? [Chiao & Cronin, 2000] How?? [Barnard, 1998] [Forsyth, 1990] [Finlayson, 1995] [Finlayson et al., 2001] [Miller & Tieu, 2001] [Schaefer et al., 2005] [Matsuda, 1995] [Tokumaru et al., 2002] [Cohen-Or et al., 2006] [Oliva & Synchs, 2000] [Oliva & Torralba, 2006] YES! important for scene recognition 3 channels account for 98% variability Mapping to illuminant (color constancy) Fixed hue intervals (color harmony) Is color really important?

11/01/06Jean-François Lalonde Natural Color Statistics p. 30 Back to psychophysics Observation:  H,S,V variation  retinal stimuli = highly non-linear! Hypothesis:  Retinal stimuli is more sensible to more likely colors  Probabilistic approach [Long et al., 2006]

11/01/06Jean-François Lalonde Natural Color Statistics p. 31 Back to psychophysics [Long et al., 2006] 1600 natural images

11/01/06Jean-François Lalonde Natural Color Statistics p. 32 Our neurons match the data! [Long et al., 2006]

11/01/06Jean-François Lalonde Natural Color Statistics p. 33 I was just kidding, but now I’m really done Can natural color be compactly represented? [Chiao & Cronin, 2000] How?? [Barnard, 1998] [Forsyth, 1990] [Finlayson, 1995] [Finlayson et al., 2001] [Miller & Tieu, 2001] [Schaefer et al., 2005] [Matsuda, 1995] [Tokumaru et al., 2002] [Cohen-Or et al., 2006] [Long et al., 2006] [Oliva & Synchs, 2000] [Oliva & Torralba, 2006] YES! important for scene recognition 3 channels account for 98% variability Mapping to illuminant (color constancy) Fixed hue intervals (color harmony) Probabilistic? Is color really important?

11/01/06Jean-François Lalonde Natural Color Statistics p. 34 Conclusion What about spatial information?  See website...  s/colourPerception/colourPerception.html

11/01/06Jean-François Lalonde Natural Color Statistics p. 35 Thank you! Your reward: You’ve just read ~10 papers in 1 hour!

11/01/06Jean-François Lalonde Natural Color Statistics p. 36 References  [Barnard, 1998] K. Barnard. Color constancy overview, 1998  [Chiao & Cronin, 2000] C.-C. Chiao and T. W. Cronin. Color signals in natural scenes: characteristics of reflectance spectra and effects of natural illumination. J. Opt. Soc. Am. A, 17(2),  [Cohen-Or et al., 2006] D. Cohen-Or, O. Sorkine, R. Gal, T. Leyvand, and Y.-Q. Xu. Color Harmonization. SIGGRAPH, 2006  [Finlayson, 1995] G.D. Finlayson. Coefficient color constancy. Ph.D. Thesis, Simon Fraser University, School of Computing, 1995  [Finlayson et al., 2001] G.D. Finlayson, S. Hordley and P. Hubel. Color by correlation: A simple, unifying framework for color constancy, PAMI 23(11) 2001  [Forsyth, 1990] D. A. Forsyth. A novel algorithm for color constancy. IJCV, 5(1):5-36, 1990  [Long et al., 2006] F. Long, Z. Yand, and D. Purves. Spectral statistics in natural scenes predict hue, saturation and brightness. Proc. Natl. Acad. Sci, 103(15), 2006  [Matsuda, 1995] Y. Matsuda. Color design, Asakura Shoten, 1995  [Miller & Tieu, 2001] E. Learned-Miller and K. Tieu. Color Eigenflows: Statistical Modeling of Joint Color Changes. ICCV, 2001  [Oliva & Synchs, 2000] A. Oliva and P.G. Schyns. Diagnostic colors mediate scene recognition. Cognitive Psychology, 41, 2000  [Oliva & Torralba, 2006] A. Oliva and A. Torralba. Building the Gist of a Scene: The Role of Global Image Features in Recognition. Progress in Brain Research: Visual perception, 155, 23-36, 2006  [Schaefer et al., 2005] G. Schaefer, S. Hordley, and G. Finlayson. A combined physical and statistical approach to colour constancy. CVPR, 2005  [Tokumaru et al., 2002] M. Tokumaru, N. Muranaka, and S. Imanishi. Color design support sustem considering color harmony. International Conference on Fuzzy Systems, 2002.