SLIDE 1 Photographic Tone Reproduction for Digital Images
Paper by: Erik Reinhard, Michael Stark, Peter Shirley, and James Ferwerda
SLIDE 2
Tone Reproduction Problem
How should we map measured/simulated scene luminances to display luminances and produce a satisfactory image?
SLIDE 3
SLIDE 4 Key Ideas
- Zone
- Middle-grey
- Dynamic range
- Key
- Dodging-and-
burning
SLIDE 5
SLIDE 6 Algorithm
- Apply luminance mapping
- If necessary, apply automatic dodging-
and-burning
SLIDE 7 Initial Luminance Mapping
SLIDE 8
SLIDE 9
SLIDE 10
Automatic Dodging-and-burning
Seek first scale sm where:
SLIDE 11
SLIDE 12
SLIDE 13 Global vs. Local operator
Global Local
SLIDE 14
Results
SLIDE 15
SLIDE 16
SLIDE 17
Discussion
Questions?
SLIDE 18
Question 1
What is problematic in the algorithm?
SLIDE 19 Question 1
What is problematic in the algorithm?
Set parameters
“Sharpening” Phi
Scale alpha_1, alpha_2
Threshold epsilon
Scale s
Key value a
“Magic”
SLIDE 20
Question 2
Why does equation 9 increase local contrast?
SLIDE 21
Question 3
What do we change in the algorithm to obtain images like this?
SLIDE 22
Question 3
What do we change in the algorithm to obtain images like this? Make histogram peak in high/low tone area → key value a
SLIDE 23
Question 4
Why can we obtain luminance like this? L = 0.27R + 0.67G + 0.06B
SLIDE 24
Question 4
Why can we obtain luminance like this? L = 0.27R + 0.67G + 0.06B
SLIDE 25 Question 4
Why can we obtain luminance like this? L = 0.27R + 0.67G + 0.06B → Luminosity function (see Wikipedia) → Scotopic vs. photopic vision Scotopic, sensitivity of the
eye is mediated by rods
SLIDE 26
Question 5
Why is it called dodging & burning? Why does burning darken the image? Why does dodging lighten it?
SLIDE 27
Question 5
Darkroom → prints are made from negatives → negative process: more light → darker