Managing the storage and battery resources in an image capture device digital camera using dynamic t

1
Managing the Storage and Battery Resources in an
Image Capture Device (Digital Camera) using
Dynamic Transcoding
Surendar Chandra Carla Schlatter Ellis Amin Vahdat

surendar_at_cs.uga.edu,carla,vahdat_at_cs.duke.edu htt
p//www.cs.duke.edu/surendar/research/
2
Problem Scenario Amateur Photographer

• Big question Which will run out first?
• His desire to take yet another picture?
• Storage space in the camera?
• Battery?

3
Problem Scenario Serious Amateur
How can she store the highest quality
shots? Constraints memory battery connectivity?
4
Digital Camera Architecture
Wireless Network (not available in typical
cameras)
Optical
Storage
Battery power
5
Digital Camera Components

• Optical
• Image capture
• device CCD, CMOS
• resolution presently upto 6 megapixel
• 8x10 print needs 40 true megapixel
• higher resolution larger file size
• View Finder
• Full LCD, traditional
• Storage
• Floppy, flash, microdisk, CDR
• Battery
• Rechargeable, primary

6
Image Quality Levels (transcoding)
7
Why reduce Quality?

• Why buy a high quality camera and lower quality?
  
• Why not just buy more memory?
• Memory is expensive for normal amateur use
• Can often get large file savings for small
  degradation in quality
• Not all images are "important"
• Need high quality camera to capture that classic
  picture
• No a-priori knowledge of the classic picture
• capture images in high quality
• preview image
• classify importance

8
Motivation

• Proliferation of mobile multimedia devices
• mp3 player, digital still and video cameras ..
• 1 billion/1.8 million digital cameras in 1999
• Constrained resources
• storage, battery, network resources
• Goal
• Explore quality vs storage vs battery power
  tradeoff

9
Research Goal

• Storage effects
• Dynamic transcoding and storage space
• Battery power effects
• Dynamic transcoding and energy consumption
• Wireless network effects
• storage
• energy consumption

10
Space Reclamation Policies

• No transcoding
• Always transcode
• current practice, capture image at final Quality
  Factor
• One shot transcode
• when necessary, transcode to final Quality Factor
• Gradual transcode
• reduce in steps of 25 till enough space reclaimed
• Upload Image

11
Performance Metrics

• Number of images stored
• Quality Factors after capturing n images
• Energy consumed after capturing n images

12
Implementation Details

• Developed simulator to capture n images
• Simulator parameters
• Processor (Hitachi SH-4)
• Storage (flash and microdisk)
• Network (Rangelan and CDPD)
• Battery (NiMH and Lithium)
• Optical unit (const. battery power consumption)
• Image workload (2.1 megapixel cameras)

13
Workload

• 30 high quality images from 2.1 megapixel cameras

Most images smaller than 1.5 MB
14
Storage Constraints (8 MB)
Higher image Quality for more images is preferred
No transcode 8, Always - 0, One shot - 10,
Gradual 32
15
Energy Consumption (microdisk)
Always transcode - 83, One shot - 80, Gradual 75
16
Power components (microdisk w/o optical)
Disk spinup, write and erase consume significant
power
17
Rangelan2 (fast 120 KB/s)
Always - 83, One shot - 80, Gradual 75, Network
59
18
Rangelan2 (slow 35 KB/s)
Always - 83, One shot - 80, Gradual 75, Network
35
19
CDPD Connectivity (2.4 KB/s)
Always - 83, One shot - 80, Gradual 75, Network
8
20
Discussion

• LCD viewer, CCD etc. drain significant power
• emerging hardware technologies
• dynamic transcoding technologies can provide
  better image quality
• energy characteristics of flash / microdisk
  (erase/disk spinup etc) important
• wireless network connectivity feasible
• network speed affects battery power consumption
• important to expose the quality vs size vs
  battery tradeoffs to the user

21
Image transcoding
Effective when lose less in quality than gain in
space saving
Low JPEG Quality 10 KB
Grayscale 85KB
Original 116 KB
Crop 40KB
Foggy road with fall foliage canopy
Thumbnail 2KB
Transliterate few bytes
22
Context

• USITS99 - JPEG Compression Metric as a Quality
  Aware Image Transcoding
• WOWMOM99 - Multimedia Web Services for Mobile
  Clients Using Quality Aware Transcoding
• INFOCOM00 - Application-Level Differentiated
  Multimedia Web Services Using Quality Aware
  Transcoding

23
Proliferation of digital cameras

• 1 billion sales/1.8 million units
• 61 use four or more hours/wk
• 83 share pictures via e-mail
• 38 post photographs on web sites
• 70 process photographs
• NPD market study (for 1999)
• NPD Intelect usage study (Oct 99)

24
Sample Interface
Net enabled lower Q
Net disabled lower Q