"Investigating Decreasing Energy Usage in Mobile Apps via Indistinguishable Color Changes" by Tedis Agolli, Lori Pollock, James Clause MobileSoft'17 Buenos Aires, Argentina, 2017.

1. Investigating Decreasing Energy Usage in
Mobile Apps via Indistinguishable Color
Changes
University of Delaware
Tedis Agolli, Lori Pollock, James Clause

2. We hate low batteries!
 Apps are becoming more powerful
 Resources are constantly in demand
 Battery development has hit a ceiling

3. Dong et al.
Power-saving color
transformation
2009
2015
Timeline
Saving Energy via UI Optimization
Li et al.
Nyx: A display energy
optimizer for mobile
web apps
Linares-Vasquez et al.
A multiobjective approach
on optimizing GUIs
Wan et al.
Detecting Display Energy
Hotspots in Android Apps

4. Optimizing the User Interface of Mobile Apps
 OLED screens consume less energy than LCD screens
 OLED energy depends on colors displayed

5. Prior to our work
Original Image Energy Optimized Image
- Energy optimized version
achieves up to 60% energy
saving
- Savings come at the cost of
heavily altering the UI
Our Goal:
Explore potential energy savings of changing UI color, while
maintaining UI aesthetics.

6. Research Plan
Conduct empirical study of mobile apps.
1. Impact. What are the energy impacts of indistinguishable color
changes?
2. Suitability. Are users able to detect differences between the
original and optimized UIs?
ExperimentVariables: Methodology:
1. Phone Displays 1. Execution Optimization
2. Subject Executions 2. Data Collection

7. Phone Displays
2.83 inch µOLED-
(AMOLED)Samsung Galaxy S2 Samsung Galaxy Nexus
- Devices used byWan et al. in their methodology
- Using the same devices allows us to compare results

8. Subject Executions
Creating Executions:
1. Manually execute app
2. When UI changes, capture a
screenshot and store as:
<screenshot, timestamp>
3. Compute duration of each
screenshot
• Duration (s) –Total duration of
execution
• # Screenshots –Total number
of screenshots
Application Duration (s) # Screenshots
Facebook 554 116
Facebook
Messenger
268 55
FaceQ 470 96
Flashlight 51 20
Instagram 429 93
Pandora 278 75
Skype 254 35
Snapchat 465 142
Twitter 388 101
WhatsApp 242 65
- Top 10 applications in Play Store
- Used byWan et al

9. Methodology
■ Input:
-Execution
-Display
Steps:
1. Build color map for image
2. For each color, find a new color that is:
a) Indistinguishable
b) Cheaper to display
3. Create new image by replacing colors
4. Measure power expenditure of old and
new image
CIEDE 2000 Color Difference Algorithm
Measure how different 2 colors are.
-Taking into account human perception
If α <= 2.3
- There is a Just Noticeable Difference
(JND)

10. Research Questions
1. Impact. What are the energy impacts of
indistinguishable color changes?
2. Suitability. Are users able to detect differences
between the original and optimized UIs?

11. RQ1: Impact
For indistinguishable changes, savings are:
- AMOLED: 3.3% - 7.8%, Mean=7.1%
- S2: 0.9% - 4.9%, Mean=4.1%
- Nexus: 0% - 0%

12. RQ2: Suitability
Survey
■ 3 Participants
■ 50 pairs of screenshots (40 + 10 test)
■ OnlineQualtrics survey
Steps:
1. Participants are shown two images side by side.
2. Respond to the statement:
“The colors in these images are effectively the same.” using
a likert scale.
(Strongly Agree, Agree, Disagree,Strongly Disagree)

13. Results:
14 28 73 5
Strongly Disagree Strongly Agree
■ 65% of answers are Agree / Strongly Agree
■ The majority of answers state that the images are effectively the
same
“The colors in these images are effectively the same.”

14. Conclusions
■ Study of energy impacts of indistinguishable UI changes
■ Results suggest:
– Moderate savings can be achieved. (Max 7.8%)
– Results of the changes are considered effectively the same.

15. AMOLED Nexus

17. Background
■ Detecting Display Energy Hotspots in Android Apps (Wan et al.)
– Built 3 power models
– Explore energy saving by modifying UI
– Perform evaluation on 10 mobile apps