Chameleon: A Color-Adaptive Web Browser for Mobile OLED Displays
Mian Dong and Lin Zhong Rice University
Chameleon: A Color-Adaptive Web Browser for Mobile OLED Displays - - PowerPoint PPT Presentation
Chameleon: A Color-Adaptive Web Browser for Mobile OLED Displays Mian Dong and Lin Zhong Rice University Display is a major power consumer in a smartphone A. Carroll, "A analysis of power consumption in a smartphone, 2010 O rganic L
Chameleon: A Color-Adaptive Web Browser for Mobile OLED Displays
Mian Dong and Lin Zhong Rice University
is a major power consumer in a smartphone
power consumption in a smartphone,“ 2010
2010 2009 2011
OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!!
OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!! OLED Rocks !!!
An OLED-friendly theme works for
but not for
in the web are White
LiveLab: A field study (25 users; 12 months)
solves the problem?
Non-Mobile Mobile OLED- Friendly
visited by iPhone users are
LiveLab: A field study (25 users; 12 months)
Max: 70% Min: 20% Median: 50%
1
Generate Device Specific OLED Power Model
i = 1 N N i = 1
Google Nexus One Nokia N85 Samsung Galaxy S
6 255
R G B
255
R G B
255
R G B
Linear RGB Values Power (μW)
2
Treat GUI Objects and Images Differently
GUI Objects vs. Images
Color Transformation of GUI Objects
RGB Pixel # RGB Pixel #
Color Counting Color Mapping Color Painting
Color Transformation of Images
3
Keep Color Consistency for Each Website
Color Consistency per Website
LiveLab: A field study (25 iPhone users; 12 months)
0% 20% 40% 60% 80% 100% Top 1 Top 5 Top 10 Top 20 Average % of usage (with Max and Min) of all users
4
Calculate Color Maps Offline
Color Transformation of GUI Objects
RGB Pixel # RGB Pixel #
Color Counting Color Mapping Color Painting
50% 60% 70% 80% 90% 1 2 3 4 5 6 7 8 9 10 11 12
Display Power Reduction Week
Optimal Trained by 1w Trained by 2w Trained by 3w Trained by 4w
2 Weeks of training work for 3 Months
http://confabulator.blogspot.com/2007/01/how-little-web-sites-have-changed-over.html
5
Give User Options
20 Participants
Dark
R’ = λR G’ = λG B’ = λB
↓25%
Green
R’ = λRR G’ = λGG B’ = λBB
↓34%
Arbitrary
R’ = R* G’ = G* B’ = B*
↓72%
Inversion
R’= λ(1-R) G’= λ(1-G) B’= λ(1-B)
↓66%
Original
2 4 6 8 CNN Facebook Google Weather ESPN User Number Inversion Arbitrary Green Dark
Different users prefer different transformations for a website
1 2 3 4 5 CNN Facebook Google Weather ESPN Scores (1 to 5) Inversion Arbitrary Green Dark
Even the same user may favor different color transformations for different websites
1
Generate Device Specific OLED Power Model
2
Treat GUI Objects and Images Differently
3
Keep Color Consistency for Each Website
4
Calculate Color Maps Offline
5
Give User Options
Model Building Color Mapping Color Counting Color Painting
Early Stages Layout Calculation Painting Display
Mobile Device
Chameleon Service Chameleon Browser Engine
Color Maps Power Model Color Histogram
Render Tree w/ Layout Bitmap
Mobile User User Options
Color Counting
RGB Pixel #
Resource Loading Parsing Style Formatting Layout Calculation Internet Scripting Painting Display DOM Tree Render Tree Render Tree w/ Layout Bitmap
Mapping Optimization
– Arbitrary
min Power s.t. for any i, j ∆E ((Li’, ai’, bi’), (Lj’, aj’, bj’)) = λ∙∆E ((Li, ai, bi), (Lj, aj, bj))
RGB Pixel #
L a b
Painting GUI Objects
DrawPoint (x, y, RGB) DrawPoint (x, y, LUT(RGB))
Resource Loading Parsing Style Formatting Layout Calculation Internet Scripting Painting Display
Implementation
Chameleon Fennec
A Field Trial
36 Participants; 3 Months
Female Male Age <20 Age 26-30 Age >30 Age 21-25 Art & Literature Business Computer Science Engineering Education High School Students Law Science
Transformation is Well Accepted especially with a Low battery level
Original Transformed
Battery Level High Battery Level Low
Summary
system power reduction for web browsing
users if performed properly
studying users