Rochester Institute of Technology RIT Scholar Works

Presentations and other scholarship

11-12-2002 Meet iCAM: A next-generation appearance model Mark Fairchild

Garrett ohnsonJ

Follow this and additional works at: http://scholarworks.rit.edu/other

Recommended Citation Fairchild, Mark and Johnson, Garrett, "Meet iCAM: A next-generation " (2002). Accessed from http://scholarworks.rit.edu/other/106

This Presentation is brought to you for free and open access by RIT Scholar Works. It has been accepted for inclusion in Presentations and other scholarship by an authorized administrator of RIT Scholar Works. For more information, please contact [email protected]. Meet iCAM: A Next-Generation Why Are We Here? Color Appearance Model

CIC X, 2002

Mark D. Fairchild & Garrett M. Johnson • Spatial, Temporal, & Image Quality Questions Remain RIT Munsell Color Science Laboratory www.cis.rit.edu/mcsl • E.g. Pattanaik et al. 1998

Outline Very Brief History of Color Appearance Models • Very Brief History of Color Appearance Models • You’ve just heard two talks on CIECAM02 • Image Appearance Modeling • iCAM: An Image Appearance Model • Enough said… • Future Directions

1 What Does a Color Appearance Model Enable? Appearance Correlates

• , • Mapping from Measurements to Words • , Chroma, Saturation (Physics to ) •

• Prediction of Color Matches (or Changes) across Changes in Viewing Conditions

History of Color Appearance Models A Next Generation CAM

1970’s: CIELAB and CIELUV

Early 1980’s: Initial Hunt and Nayatani Color Appearance Models

Late 1980’s: Revisions of Hunt and Nayatani Models

Early 1990’s: Model Testing, Further Revisions, New Models (e.g., RLAB, LLAB)

Late 1990’s: Convergence … CIECAM97s

Early 2000’s: Widespread focused testing and refinement, CIECAM02 is an Evolution of CIECAM97s CIECAM02, Practical Image Appearance Models New Capabilities Require a New Approach …

2 What is an Image Appearance Model? What are Some of the Missing Links? • Spatial Vision (Filtering & Adaptation) • Image appearance models extend color • Scene Interpretation appearance models to include spatial vision, • Computational Surround Effects temporal vision, and image difference/quality • Color/Image Difference Metrics properties. • Image Processing Efficiencies

• They account for more complex changes in visual response in a more automated manner.

scene representation

spatial sampling, optical image formation

simulated retinal image

calculation of photoreceptor responses MOM Rod S-cone M-cone L-cone Image Quality Measurement response response response response image image image image rod S M L

pyramidal image decomposition

lowpass lowpass lowpass lowpass Gaussian Gaussian Gaussian Gaussian pyramid pyramid pyramid pyramid rod S M L upscaling, subtraction Johnson & Fairchild, CIC (2001) bandpass bandpass bandpass bandpass contrast contrast contrast images images images images Pattanaik, Ferwerda, Fairchild, & rod S M L band-limited, local, adaptation processing Visual Encoding

adapted adapted adapted adapted contrast contrast contrast contrast signals signals signals signals Greenberg, SIGGRAPH and CIC (1998) rod S M L opponent transformation

adapted adapted adapted adapted contrast contrast contrast contrast signals signals signals signals rod A C1 C2 Modular Framework for IQ Scales nonlinear contrast transduction, thresholding

perceived perceived perceived contrast contrast contrast signals signals signals A C1 C2 Multi-Scale Observer Model inversion of visual encoding for display •Promising Framework for Image (mapped) (mapped) (mapped) bandpass bandpass bandpass contrast contrast contrast images S images M images L image reconstruction Differences Comprehensive (mapped) (mapped) (mapped) S-cone M-cone L-cone response response response image S image M image L

color space transformation, gamma correction Display Mapping •Flexible Implementation displayed displayed displayed Extremely Complex image image image R G B Successfully Implemented 2-3 Times!!

3 Image Difference Process Meet iCAM Reproduction 1 Reproduction 2 iCAM — image Color Appearance Model A simple framework for color appearance, spatial vision effects, image difference (quality), image processing, and temporal effects (eventually).

* * Mean DE ab 2.5 Mean DE ab 1.25

Mean DIm 0.5 Mean DIm 1.5

Spatial Filtering, Local Attention, Local & Global Contrast, CIE

Pointwise iCAM Spatial iCAM

4 iCAM Performance Examples Basic Appearance Attributes

• Transform (CAT) • Chromatic Adaptation Transform (CAT) – Identical to CIECAM02 • Color Appearance Scales • Constant Hue Lines • Color Appearance Scales • Simultaneous Contrast – Similar to Munsell / CIECAM02 (limited) • Chroma Crispening • Hue Spreading • Constant Hue Lines • HDR – Best Available (IPT) – Facilitates Mapping • Image Difference (Quality)

iCAM Simultaneous Contrast iCAM Chroma Crispening

Original Image iCAM Lightness Original Image iCAM Chroma

http://www.hpl.hp.com/personal/Nathan_Moroney/

5 iCAM Spreading iCAM High-Dynamic-Range Tone Mapping

Earlier-Model Results Original Image iCAM Hue www.debevec.org

iCAM Image Difference (Image Quality) iCAM Image Difference (Image Quality)

(a) 14 1.2

12 1

10 0.8 8 DIm 0.6 DIm 6

0.4 Model Prediction

Model Prediction 4

0.2 2

0 0 -5 -4 -3 -2 -1 0 1 2 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 Perceived Difference Perceived Contrast

Image Difference Prediction (Sharpness Data) Image Difference Prediction (Contrast Data)

6 Image Input Data Transform to “Sharpened Cone Responses”

XYZadapt: Different Filters for luminance & TBD The same 3x3 as adopted for CIECAM02. All Filters: Viewing-Distance Dependent

TC8-01 (CIECAM02) Linear CAT Surround & Luminance Dependent Transform to IPT Color Space

The same CAT as CIECAM02 for simple conditions. “Contrast” depends on luminance and surround.

7 Rectangular-to-Cylindrical Conversion to Brightness, Colorfulness Lightness, Chroma, Hue

Just plain geometry. Dependency on absolute luminance.

Spatial iCAM Mathematica Notebooks

Detailed references to each step in proceedings.

Coded examples on the internet. • Can be read and executed with free “MathReader”. Open-Source Science • Other code (Matlab, IDL) forthcoming. • Updates

8 Conclusions Future Directions

• iCAM Represents an Example of a New Generation of • HDR Digital Photography (Capture & Processing) Color Appearance Model • Video iCAM (Temporal Adaptation & Filtering) • HDR Digital Video (Processing) • Image Appearance Model • Better Understanding of Surround Effects • Incorporates Spatial Vision • Image-Content Dependent Reproduction • Can Be Extended for Temporal Vision (EI 2003) • Refined Image Difference & Image Quality Metrics • Image Difference Metric, DIm (EI 2003) • Extension to Preferred Image Reproduction • Basis for a Fundamental Image Quality Metric • Psychophysics, Psychophysics, Psychophysics

Suggestions and Help Welcome and Encouraged

9