1. Development of Softcopy Environment for Primary Color Banding Visibility Assessment
Byungseok Min*, Zygmunt Pizlo**, and Jan Allebach*
*School of Electrical and Computer Engineering
**Department of Psychological Sciences
Purdue University, West Lafayette, IN
Research supported by Hewlett-Packard Company

2. Outline
Motivation
Development of hardcopy environment for banding discrimination
Development of softcopy environment
Verification of the softcopy environment
Conclusion

3. Past Studies on Banding Assessment
Cui et al.(2001): measurement of monochrome banding threshold in inkjet printers
Bang et al.(2003): monochrome banding discrimination in hardcopy environment
Arslan et al.(2005): monochrome banding discrimination in softcopy environment
Chen et al.(2004): softcopy environment for sinusoidal artificial banding of
secondary color

4. Motivation 1-D space  beginning with primary colorant
Hardcopy Image with intrinsic banding
Paper process direction
Softcopy Image with intrinsic banding
Constant image
OPC drum velocity variation
Why we need softcopy ?
A color is represented by 3-D signal, but the banding perception is described in
1-D space  beginning with primary colorant
The same banding perception in both hardcopy and softcopy environment 4

5. Generation of Extrinsic Banding in Hardcopy Environment
1-D representation
Extracting banding profile
Intrinsic
banding
image
Line
pattern
image
1-D Banding profile
pixel
(1~200 cpi)
Printer
Scanner
Curve
Projection
& Filtering dE
Scanner
Banding Level
Random phase
Printer Σ
Extrinsic
Banding
Image
Extrinsic banding level
PWM coding

6. 1-D Representation of Color Banding Signal
Test patch
Scanned patch
Printer
process direction
Scanner
Line pattern
(PWM code)
Line pattern definition
Scanner
calibration
projected point of * *
if Yk > Yavg *
projected point of
if Yk < Yavg *

7. How Do We Print Extrinsic Banding ?85 83 84 87
84.5
83.5 22 63 20 21 24 22 63 20 21 24 ….
paper process direction
PWM line code
1-D banding profile(ΔC)
PWM index
Extrinsic banding
Exploit sub-pixel addressability of the laser beam in the scan direction to perform PWM

8. How pulse-width modulation (PWM) works with laser electrophotography
Normally, we think of a monochrome laser EP printer as rendering a binary image, in which each pixel is either black or white.
However, as suggested by the architecture diagram shown on the right, we can switch the laser beam on and off for a fraction of the width of each pixel in the scan direction.
For example, we can assign an 8-bit PWM code to each pixel. The most significant 2 bits determine the justification mode (left, centered, right, and split). And the 6 least significant bits provide 64 sub-pixel pulse widths (0 = no pulse, 63 = full width pulse). The figure to the right does not include the center-justification mode.

9. Outline Motivation Development of hardcopy environment
Development of softcopy environment
Haploscopic matching method
Appearance mapping function in adapted XYZ space
Verification of softcopy environment
Conclusion

10. Softcopy Environment Overview
Extrinsic banding generation
1-D representation
Extrinsic
Banding
Image
Scanned
Patch
Background
Image + + -
Scanner
Curve
Projection
& Filtering
Back-
projection Σ Σ +
Back-
projection
Extrinsic
Banding
Image
Brief explanation
Adapted Space of Observer
Banding Level β
Color
Appearance
Mapping
Function
Inverse
Monitor
Curve
Spectro-
radiometer
Calibration
Color appearance matching process

11. Appearance Matching Experiment (Haploscopic Matching Setup)
Front view
Lighted room
(Fluorescent)
Viewing Box
(Fluorescent)
Memory matching method works with monochrome
Dark surround
Independent adaptation of each eye for each viewing condition 11

12. Another view of haploscopic matching set-up

13. Appearance Matching Procedure
Fine adjustment
Gray adaptation (1 min)
Rough selection
Observe hardcopy with left eye and softcopy with right eye
Adjust a color by changing hue, chroma, and lightness

14. Device Characterization
Scanned image
XYZX-Rite
XYZspectro
XYZsoftcopy
Scanner calibration
Appearance mapping function
Viewing booth
Spectrophotometer
Barco monitor
Spectroradiometer
Mapping function
Delta E
Avg: 0.38
Max: 1.33

15. Adapted XYZ space Viewing condition 1 Lighted surround / paper
Von-Kries adaptation model
XH, YH, ZH
XHa, YHa, ZHa
Measured by
spectro-
radiometer
Mapping function
Xsa, Ysa, Zsa
Xs, Ys, Zs *
Viewing condition 2
Dark surround / monitor
* : Hunt-Pointer-Estevez transformation

16. Appearance Mapping Function
Experiment result
Appearance mapping function
Fitting error
Avg deltaE: 2.13
Max deltaE : 2.98

17. Outline Motivation Development of hardcopy environment
Development of softcopy environment
Verification of softcopy environment
Banding level matching experiment
Discrimination threshold (DL) experiment
Conclusion

18. Softcopy Environment Validation 1: Banding Level Matching
Hardcopy (Lighted surround)
Softcopy (Dark surround)
3 different levels of banding are matched by memory matching method
Memorize the banding level of the hardcopy in the lighted surround
Match the banding level on the monitor in the dark surround

19. Banding Matching Experiment Result
Total 12 subjects participated in the experiment
For each subject, three pairs of
Linear Regression model
Hypothesis
- Slope should be statistically equal to 1.0 in 95% confidence interval
- Intercept should be statistically equal to 0
Result
+ intercept
Parameter
Standard error
95% lower limit
95% upper limit
Slope
0.088
0.988
1.345
Intercept
0.036
0.009
0.165
The interval includes 1.0
The Intercept is almost 0

20. Softcopy Environment Validation 2: Difference Threshold (DL) Comparison
Hardcopy (Lighted surround)
GUI
Softcopy (Dark surround)
To remove individual variability
The same experiment is performed in both environment
Method of constant stimuli is used
Probit analysis is used to estimate DL
Environment
Hardcopy
Softcopy
Average DL
0.103
0.091
95% C.I.
± 0.028
±0.016

21. Discrimination Experiment Result
Two-way ANOVA analysis
- Two factors: Subject(# of subjects) and Environment (Hard/Softcopy)
Hypothesis
- H0: The environment factor does not affect the result
- The null hypothesis is rejected if P-value is less than 0.05
Result by SAS program
Source DF
Mean Square
F-value
P-value
Environment 1
0.60
0.4546
Subject 11
1.41
0.2903
No difference
- between two environments
- among 12 subjects

22. Conclusion and Future Work
We Successfully developed and verified our softcopy environment for banding discrimination
DL of 0.09 suggests that the banding magnitude should be reduced by 9% to yield noticeable reduction in the banding visibility
Future work
Color banding assessment on secondary colors

23. Question ?