A Computational Darkroom for BW Photography

1
A Computational Darkroom for BW Photography

• Soonmin Bae, Sylvain Paris,
• and Frédo Durand
• Current Status Resubmission to Siggraph

2
Objectives

• To enhance black-and-white photographs
• Look transfer between two images
• Direct interpolation and manipulation of the
  look

3
What can be the Look
4
Approaches

• Decomposition of an image into large-scale
  variation layer and high frequency texture layer
• Control the global contrast and the local
  textureness separately
• Quantitative characterization
• Use image statistics and histograms

5
To Do

• Control of the visual quality, look
• Parametric characterization
• User-oriented and intuitive method
• HDR images

6
Not To Do

• Deal with
• Color photographs
• Paintings
• Change Content
• Change Composition
• Crop
• Select a model or ideal parameters

7
What they do vs. What we do

• Tone mapping
• Ferwerda et al. 1996Tumblin and Rushmeier 1993
  Ward 1994
• Ashikhmin 2002 Tumblin and Turk 1999 Pattanaik
  et al. 1998 Reinhard et al. 2002
• Color2gray
• Gooch et al. 2005
• Image analogies
• Hertzmann et al. 2001 Efros and Freeman 2001
  Rosales et al. 2003 Drori et al. 2003

• Objective tone reproduction vs. Control of the
  look
• Non-parametric vs. Parametric characterization

8
Challenges

• Identification of important visual
  characteristics
• Meaningful feature selection
• Decomposition
• Faithful extraction of the features
• Reconstruction
• Halo artifact
• Subjective issues
• Preference vs. Similarity

9
Technical details - Separation

• Blah.. Blah..

10
Preliminary Results
11
Expected Demos

• Blah.. Blah..

12
Open Discussion

• Should we include the following domains?
• Color photographs
• Paintings
• Which should be pursued?
• Transfer vs. Direct parameter modification
• Similarity vs. Preference