projects logistics

1

Tone Reproduction

Assignments

• Checkpoint 6 – Due Monday

• Checkpoint 7– To be given Monday

• RenderMan – Due Nov 3rd

Projects

• Approx 22 projects• Listing of projects now on Web• Presentation schedule

– Haven’t scheduled? Please do so!

Logistics

• Final Report– Introduction – Approach Taken– Implementation Details– Results– Appendix/Code

• All project material due Friday, Nov 19th

– No late submission• else I can’t get your grades in!

This and Next Week

• Tone Reproduction Week!– Today

• Intro to tone Reproduction– Erik Reinhard (U of Central Fla)

• Colloquium – Monday, Nov 1st / 1-2 pm /70-1400 (auditorium)

• Guest Lecture (Tone Reproduction)– CGII Class

– Joe Geigel• Colloquium

– Thursday, Nov 4th / 1-2pm / 70-3000 (CS conf room)

Computer Graphics as Virtual Photography

camera (captures light)

synthetic image

camera model

(focuses simulated lighting)

processing

photo processing

tone reproduction

real scene

3D models

Photography:

Computer Graphics:

Photographic print

2

Tone/Color Reproduction

• Where are we?– Described our scene during modeling– Simulated light transport during rendering– Captured and projected light from the scene

onto a 2D plane during capture– Now we must convert this simulated light

capture into an image for display

Tone Reproduction

• Luminance levels

Sky = 12400 nits

Trees = 64 nits

Traditional Photography

camera

processing

photo processing

real scene

Photographic printPhotography:

Reinterpretation of scene optimized for viewing

Digital Photography

camera

processing

Processing performed by camera

real scene Digital image

Photography:


(24 bit RGB)

Digital Photography

• Issues–Tone Reproduction is “hard coded”

into camera–Color Management Issues

• Which RGB?• Optimized for what display?

Image Synthesis in CG

camera

synthetic image

camera model

processing

photo processing

tone reproduction

real scene

3D models

Photographic printPhotography:

Computer Graphics:


(24 bit RGB)

Scene luminance

3

High Dynamic Range (HDR) Imaging

• high dynamic range imaging is a set of techniques that allow a far greater dynamic range of exposures than normal digital imaging techniques.

• The intention is to accurately represent the wide range of intensity levels found in real scenes, ranging from direct sunlight to the deepest shadows.

Wikipedia

HDR in Computer Graphics

[Ward 2001]


[Debevec 2001]


[Debevec 2004]

4

What if we ignore tone Reproduction?

• Simple Linear tone reproduction

Light source = firefly Light source = Searchlight

[Tumblin93]

Tone Reproduction

Definition: Compressing the dynamic range of a scene’s luminances/radiances so that it can be displayed on a given device in such a way that minimizes the perceptual difference between viewing the scene and viewing the rendering of the scene.

Tone Reproduction - Definition

• Dealing with luminances / radiances• Rendering will be displayed on a given

device• Minimize perceptual difference between

real and created.

Tone Reproduction

• Radiance / Luminance– Flux arriving at or

leaving from a given point or surface in a given direction.

– Radiance measured in W / m2 /sr

– Luminance measured in cd/m2 (nit)

dA

5

Tone Reproduction

• Using 0 – 1 to indicate light intensity– What does 1 mean?

• CG tends to use intensity space of output device

• Images optimized for a given output device.

Why Tone Reproduction?

• Human response to light is neither simple nor linear.

• Most display devices are not linear• Incorrect response modeling results in

incorrect perception of results.

The Tone Reproduction Problem• What operator will create a close match between real-

world and display brightness sensation?

[Tumblin93]

Tone Reproduction in CG

[Ferwerda 1998]

Tone / Color Reproduction

• Response / Observer– How does a system (like the human visual

system or photography) respond to the collected light?

• Display– How do we translate that response using a

particular output device (like a CRT or printer)?

Response Models

• Applying observer/response model will result in the luminances as seen by your display observer.– i.e., will be in luminance range of your output

device.• Observer/Response Models

– Human Visual System (today)– Photographic Systems (Monday)

6

Response Models

• Image Characteristics– Spectral response - how system responds to

different wavelengths of light– Intensity response - how system responds to

different intensities of light– Acuity - the sharpness of the image produced by

the system– Noise – inherent noise in the image produced

by the system

Human Visual Response

Human Visual Response• Pupil

– Regulates the amount of light that gets to the retina

• Photoreceptors– Rods

• 75 - 150 million• sensitive to 10-6 to 102 cd/m2 (low light levels)• Achromatic (detects “brightness”)

– Cones • 6 - 7 million• sensitive to 0.01 to 108 cd/m2 (high light levels)• Responsible for color vision


• Levels of Brightness Response– Scotopic (Primarily rods)

• 10-6 to 102 cd/m2

– Photopic (Primarily cones)• 0.01 to 108 cd/m2

– Mesopic (overlap!)• 0.01 to 102 cd/m2

• Both rods and cones• Little known -- active area of research


• Spectral response– Human Visual System is sensitive to light in the

wavelength range of approx. 350 - 700 nm.– Sensitivity changes dependent on illumination

level

Human Visual Response• Changes in Spectral Sensitivity

Scotopic Mesotopic Photopic

[Ferwerda96]

7

Human Visual System

• Acuity– Ability to

resolve spatial detail

• Snellen Chart– View from 20 ft away– Line 8 subtends 1 min

of visual angle– People who can read

this is said to have 20/20 vision

[Ferwerda96]

Human Visual System• Acuity also changes dependent on luminance

level

[Ferwerda96]

Human Visual System• Response at different illumination levels

[Ferwerda96]

Human Visual System

• Adaptation– Our vision system has the ability to adapt to a

given luminance level– Light Adaptation - from darkness to light– Dark Adaptation - from brightness to dark– Adaptation is gradual, not immediate (and is

subject to age! )

Human Visual System

• Threshold Studies– determine the threshold at which a person can

notice the change between a light sample given a certain background luminance.

Human Visual System

• Time course for light adaptation

For rods For cones

[Ferwerda96]

8

Human Visual System• Time course of light adaptation

[Ferwerda96]

Human Visual System

• Time course of dark adaptation

[Ferwerda96]

Human Visual System• Time course of dark adaptation

[Ferwerda96]

Human Visual System

• Ferwerda’s model– Scales luminances as to preserve perceived

contrast using psychophysical data as a guide. • Lw = mLd

– Different models for scotopic and photopic vision with slider to blend the two to simulate mesopic vision.

• m will vary dependent upon whether scene is in scotopic, photopic, or mesopic range.

• Greg Ward offers a simpler approach in Graphics Gems, IV

Ward Tone Reproduction Original Tumblin-Rushmeier operator

• Based on “brightness”, a perceptual measure of how bright humans perceive light.

9

“Normal” Linear Mapping

[Graphics Gems, IV]

Tumblin-Rushmeier Operator

[Graphics Gems, IV]

Ward Operator Results

[Graphics Gems, IV]

Human Visual System

• A good overview of CG tone reproduction operators is available from– “Tone Reproduction and Physically Based Spectral Rendering” by

Devlin et al., State of the Art Report, EUROGRAPHICS 2002.

• Note that Tone Reproduction operators are now starting to run in real time using GPU.

• Questions? Break.

Photographic Response• Print photography process

Camera Film Process

Process

Negative

PrintPaperPrinter

[Geigel97]

Optics Photographic Material

Processed Photographic

Material

Photographic Materials

• Comprised of microscopic grains of silver halide in a gelatin (emulsion)

• Latent image formed when exposed to light• Silver halide converted to metallic silver

during processing.• Converted silver results in opacity

10

Photographic Response

• Illumination Response - high level response of an emulsion to light

• Spectral Sensitivity - Response of a material to different wavelengths of light

• Acuity - Level at which material can reproduce spatial details

• Graininess - Observed variation due to grain distribution


• Sensitometry– The science of measuring the sensitivity of

photographic materials– Each characteristic has its own unique

sensitometric measure.


• A typical brightness response / characteristic curve

Log Exposure

Den

sity

I

II

III

IV

I - toeII - straight line

sectionIII - shoulderIV - area of

solarization

γ - gamma

γ

[Geigel97]

Photographic Responsegamma - slope of region II

gives contrast rangespeed - indicates sensitivity to light

S0

1

2

3

-2 0 2

γ

0

1

2

3

-2 0 2

Photographic Response Effects of film Speed

Original 100 Speed Film

400 Speed Film 800 Speed Film

[Geigel97]

Photographic Response - Gamma

Original Low Contrast

Medium Contrast High Contrast

[Geigel97]

11

Photographic ResponseSpectral Response for Three Types of Film

0

100

panchromatic

0

100

orthochromatic

0

100

300 400 500 600

blue sensitive

(Entire visible spectrum)

(Blue/Green sensitive)

(Untreated- blue/ultraviolet)

[Geigel97]

Photographic ResponseEffects of Spectral Sensitivity

Original Panchromatic Blue Sensitive

[Geigel97]

Photographic Response - Grain

∆Di = deviation of sample i from the mean

rms deviation:

A = area of scanning aperture

Selwyn Granularity:

G = (2A) σ

Indication of sample uniformity Measure of granularity

σ 1NΣ(∆Di)2 =

2

Photographic Response - Grain

[Geigel97]

Photographic Response – Acuity (Resolution)

modulation transfer function

point spread function

0

20

40

60

80

100

0 40 80 120

spatial freq. (cycles/mm)

(%)

Photographic Response - Acuity

Without MTF With MTF With MTF & Grain

[Geigel97]

12


• Observer model can mimic response of photographic systems– Reinhard (Monday) – model based on

photographic response and photographic techniques

– Geigel (Thursday) – general model on simulating response of media to light.

• Questions?

Display Models

• Need to determine the control values (RGB) needed to produce luminances calculated by observer models

Display Models

• Two Problems to be addressed by display models– Gamma

• Luminances from observer model are on a linear scale. Most display devices are non linear

– Gamut• Chromaticities calculated by observer model may

not be reproducible on a given device due to a limited color gamut.

Display Models

– Luminances from observer model are based on a linear scale.

– Most display devices are non linear.

Display Models• CRTs respond non-linearly to voltage• This non-linearity is described by gamma

• where– Ld is the actual display luminance– Ldmax is the maximum display luminance– V is the voltage [0,1]

γ)( maxVLL dd =

Display Models

• CRTs are non-linear

Sample input to monitor Graph of input

Output from monitor Graph of output

13

Display Models

• Gamma correction

Sample input to monitor Graph of input

Gamma correction Graph of Gamma correction

Output from monitor Graph of output

Display Models

• Most displays/video cards now have gamma control as part of their OS.– If we can correct so that gamma is 1.0 then, getting

using Ldmax from specs, the voltage V is given by

maxd

d

LLV =

1/γ

Display Models

• Gamut– Range of chromaticities reproducible by a

device

Display Models

• Different Devices have different gamuts

Display Models

• Perceptual color spaces– CIELAB– Distances between color values corresponds to

difference in perception– Computed from X,Y,Z values and X,Y,Z of a

reference white.

Display Models

Handling out of gamut colors

14

Display Models

• Display Models must address– Gamma / non-linearity of device– Gamut

• Usually dealt with by Color Management Systems.

Tone Reproduction

• A final word on Tone Reproduction– Recall that viewing conditions also affect

perception– TR Operator should also make modifications if

viewing conditions of world observer does not match that of display observer

– Generally included in color management systems but not Tone Reproduction operators.

Tone Reproduction

• Summary– Means of compressing dynamic range of scene to fit

that of display– Observer / Response Model

• Human Visual System• Photographic Systems

– Device Model

• Questions?

projects logistics

Documents