Color2Gray:

Salience-Preserving Color Removal

Amy Gooch
Sven Olsen
Jack Tumblin
Bruce Gooch
 New Algorithm

Color

Grayscale
Problem

Volume of displayable CIE L*a*b* Colors

QuickTime™ and a
Sorenson Video 3 decompressor
are needed to see this picture.
Isoluminant Colors

Color Grayscale
Converting to Grayscale…

• In Color Space
– Linear
– Nonlinear
• In Image Space
– Pixels (RGB)
• Using colors in the image
• Different gray for different color
– Relative difference
• Using colors in the image and their position in image space
• Colors can map to same gray…..
Traditional Methods:
Luminance Channels

Problem can not be solved by

CIE CAM
simply97switching to a different
Photoshop
spaceLAB

CIE XYZ YCrCb

Traditional Methods
• Contrast enhancement &
Gamma Correction
– Doesn’t help with isoluminant values

Photoshop Grayscale PSGray + Auto Contrast New Algorithm

Simple Linear Mapping

Axis
e
i n anc
m
Lu
Principal Component
Analysis (PCA)
Luminance
Axis
Problem with PCA

Worst case:
Isoluminant Colorwheel
Non-linear mapping
Contemporaneous Research

• Rasche et al. [2005, IEEE CG&A and EG]

Color Image Luminance Only Rasche et al.'s Method

Goals

• Dimensionality Reduction
– From tristimulus values to single channel
Loss of information

• Maintain salient features in color image

– Human perception
Relative differences

Color Illusion by
Lotto and Purves
http://www.lottolab.org
Challenge 1:
Influence of neighboring pixels
Challenge 2:
Dimension and Size Reduction
120, 120

100

-120, -120
Challenge 3:
Many Color2Gray Solutions

Original
.
.
.
Algorithm Intuition
color2gray
For first pixel

Look
at C
1 2
L1 L
+ 1 LL22
.. L1 +  L2 + 2,1
i = 1, j = 2 .
luminance For Nth pixel

Look
at C
L11 L2L+2 2,1
L1 L2
Algorithm Overview
• Convert to Perceptually Uniform Space
– CIE L*a*b*
• Initialize image, g, with L channel
• For every pixel
– Compute Luminance distance
ij
– Compute Chrominance distance

• Adjust g to incorporate both luminance and

chrominance differences
Color2Grey Algorithm

Optimization:
i+

min  (g - g ) -  
i
i j i,j
j=i-
Parameters

 Radius of neighboring pixels

 Max chrominance offset

 Map chromatic difference to increases or decreases in

luminance values
: Neighborhood Size

= 300o
 = 10

=2  = 16 = entire image

= 49o
 = 10
: Neighborhood Size

 = 16 = entire image
: Chromatic variation maps
to luminance variation


 crunch(x) =  * tanh(x/)

=5  = 10  = 25
Perceptual Distance

• Luminance Distance: Lij = Li - Lj

• Chrominance Distance: ||Cij||

Problem: ||Cij|| is unsigned

Map chromatic difference to increases or
decreases in luminance values
+b*

C2

-a* +a*

C1 Color
Space
 +b*
Color C1,2
Difference v = (cos , sin )
Space
+
- v


-a* +a*

+
-
-b*
sign(||Ci,j||) = sign(Ci,j . v)
 = 45

= 225
Photoshop Grayscale
 = 135 = 45

= 0

Grayscale
How to Combine
Chrominance and Luminance

ij Lij (Luminance)

How to Combine
Chrominance and Luminance

Lij (Luminance) if |Lij| > ||Cij||

ij
||Cij|| (Chrominance)
How to Combine
Chrominance and Luminance

Lij if |Lij| > crunch(||Cij||)

ij
crunch(||Cij||)

120, 120

100

0

crunch(x) =  * tanh(x/) -120, -120
How to Combine
Chrominance and Luminance

Lij if |Lij| > crunch(||Cij||)

ij if Cij .  ≥ 0
crunch(||Cij||)
crunch(-||Cij||) otherwise

.
.
.

Grayscale
Color2Grey Algorithm

Optimization:
i+

min i j=i- (g - g ) -  
i j i,j

If ij == L then ideal image is g

Otherwise, selectively modulated by Cij
Results
Photoshop Color2Grey
Original Color2Grey
Grey + Color
Original PhotoshopGrey Color2Grey
Color2Grey+Color
Original PhotoshopGrey Color2Grey
Original PhotoshopGrey Color2Grey
Implementation
Performance
• Image of size S x S
– O(2 S2) or O(S4) for full neighborhood case
• 12.7s 100x100 image
• 65.6s 150x150 image Athlon 64 3200 CPU
• 204.0s 200x200 image
– GPU implementation
• O(S2) ideal, really O(S3)
– 2.8s 100x100
– 9.7s 150x150 NVIDIA GeForce GT6800
– 25.7s 200x200
Future Work
• Faster
– Multiscale

• Smarter
– Remove need to specify 
• New optimization function designed to match both signed
and unsigned difference terms
– Image complexity measures

• Animations/Video
Validate "Salience Preserving"

Original PhotoshopGrey Color2Grey

Apply Contrast Attention model by Ma and Zhang 2003

Validate "Salience Preserving"

Original PhotoshopGrey Color2Grey

Thank you
www.color2gray.info
• SIGGRAPH Reviewers
• NSF
• Helen and Robert J. Piros Fellowship
• Northwestern Graphics Group
• MidGraph2004 Participants
– especially Feng Liu
• (sorry I spelled your name wrong in the acknowledgements)
Original Color2Grey Color2Grey+Color
Original Color2Grey Color2Grey+Color
Original Color2Grey Color2Grey+Color
Original PhotoshopGrey Color2Grey
Original Color2Grey Color2Grey+Color
Original PhotoshopGrey Color2Grey
Original Color2Grey Color2Grey+Color
Photoshop Grayscale
Photoshop Grayscale
 Rasche et al.

Photoshop Grayscale
Rasche et al.

Photoshop Grayscale
Rasche et al.

Photoshop Grayscale
Parameter 

= 5 = 15 = 25 = 35 = 45

= 55 = 65 = 75 = 85 = 95