1 / 40

Saving Bitrate vs. Users: Where is the Break-Even Point in Mobile Video Quality?

Saving Bitrate vs. Users: Where is the Break-Even Point in Mobile Video Quality?. ACM MM’11 Presenter: Piggy Date: 2012.05.07. Outline. Introduction Related Work User Study Result Discussion and Conclusion. Introduction. Mobile video service is getting popular

jennis
Télécharger la présentation

Saving Bitrate vs. Users: Where is the Break-Even Point in Mobile Video Quality?

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Saving Bitrate vs. Users: Where is the Break-Even Point in Mobile Video Quality? ACM MM’11 Presenter: Piggy Date: 2012.05.07

  2. Outline • Introduction • Related Work • User Study • Result • Discussion and Conclusion

  3. Introduction • Mobile video service is getting popular • Due to the development of mobile device • Minimizing video bitrate is important • Wireless networks prefer low bitrate to adapt to different bandwidth conditions • Users prefer low bitrate as most network providers normally charge for data usage • Video providers need to save costs associated with serving the content

  4. Introduction • However…… • Low video bitrate => poor video quality • Fortunately…… • Nonlinear relationship between perceived quality and video bitrate

  5. Introduction • Goal: To find the most efficient bitrate requirement that • Optimizes bandwidth usage • Maintains good user viewing experience • Lowest acceptable video quality vs. lowest quality for long term viewing

  6. Introduction • Contribution • Mapping of video bitrates to the subjective judgment of quality pleasantness • Impact of content type, video encoding parameters and user profile on mobile video viewing experience • Users’ selection processes and their criteria for the lowest pleasing quality for different content type

  7. Related Work • Users’ requirements for mobile video depends on • Social and psychological factors • Consumption model, service, user profile, context, etc… • Video quality • Spatial and temporal resolution • Quantization • Motion and texture complexity

  8. Related Work • Factors influence the reduction of bitrate • Resolution • Frame rate • Quantization • And the degradation in perceived video quality is not proportionate to the decrease in bitrate

  9. Related Work • Subjective assessment • ITU recommendation: scale-based subjective assessment • 5/9/11-sclaes • Overburdens participants • Binary choice method for assessing acceptability

  10. Related Work • Though previous works have identified the lowest acceptable quality level • They were restricted by the technology and device at that time. • Different resolution • People behaviors have changed (got used to HD quality)

  11. User Study • Equipment • iPhone 3GS with 16GB storage • Display: 480x320 pixels • H.264/AVC • Up to 1.5 Mbps, 640x480 pixels, and 30 frames per second • AAC-LC audio format • Up to 150 kbps, 48kHz

  12. User Study • Test material - 5 content types • News, music, animation, sports and movie

  13. User Study • Test material – encoding using 3 parameters • Quantization parameters (QP) • Spatial resolution (SR) • 320x240, 480x320, and 640x480 • Frame rate (FR) • Divided into 3 groups based on SR:L, M and H with each group contain 10 test clips • 30 test clips for each content type

  14. User Study • Total 150 test clips • 30x5

  15. User Study • Participants • Lounge area outside of a university library • 40 participants • Equal number of males and females • Age range: 17 ~ 35 (average = 23.2) • User profile collection • Experience of using mobile video • Preference for content types

  16. User Study • Participants’ profile

  17. User Study • Procedure • Scenario explanation • 3 steps within 20-25 mins for data collection • Participant’s profile collection • Participant randomly chose the video contents • A short interview

  18. User Study • Customized iPhone application • Participant profile collection • Content type choice • History review • Quality adjustment • Ascending • Descending

  19. User Study

  20. User Study

  21. User Study • Interview • What criteria did you use to select the desired video quality? • Is there any difference between your criteria for different content type? Why?

  22. Result • Acceptability calculation • Lower than the selected lowest acceptable clip => 0 • Otherwise => 1 • Refers to the percentage of participants accepting a video quality as the lowest quality • Binary Logistic Regression • Video encoding parameters • Content type • Viewing order • User profile

  23. Acceptability and Encoding Parameters • Different from • Content to content • Resolution to resolution • Movie is the lowest while new is the highest • The difference reduces as the resolution increases

  24. Acceptability and Encoding Parameters

  25. Acceptability and Encoding Parameters • Acceptability group • 0 – 40% should be avoided • 41 – 60% critical state • 61 – 80% can please users • 81 – 100% high user satisfaction

  26. Acceptability and Encoding Parameters • Bitrate-acceptability curves

  27. Acceptability and Encoding Parameters • Bitrate-acceptability curves

  28. Acceptability and Encoding Parameters • Bitrate-acceptability curves • High resolution needs a higher bitrate • The acceptability of “sport” rises slower than other content types • Mapping of bitrate to acceptability

  29. Influencing factors on quality Acceptability • Significant factors • Quantization parameter • Spatial resolution • Frame rate • Content type • Gender • Frequency • Duration • Viewing order • Non-significant factors • Age

  30. Influencing factors on quality Acceptability • Effect of content type • Movie vs. music, news, and animation • Spatial resolution decreases => content type more significant • Effect of encoding parameters • Video quality increases with • Decrease of QP (great difference among adjacent QP values) • Increase of SR • Increase of FR

  31. Influencing factors on quality Acceptability • Effect of viewing order • Acceptability in descending order is lower than ascending order • Significant for animation, music, news and sports but not for movie

  32. Influencing factors on quality Acceptability • Effect of user profile

  33. Influencing factors on quality Acceptability • Effect of user profile • Gender vs. frequency

  34. Influencing factors on quality Acceptability • Effect of user profile • duration vs. frequency

  35. Influencing factors on quality Acceptability • Effect of user profile • Users’ preference

  36. Quality selection patterns • Average time spent on switching is different from content type to content type • News is the lowest

  37. Quality selection patterns • Two selection patterns • Directly choose the target qualities without hesitation – mostly in ascending order • Bounced to and from the lower of higher quality for comparison – mostly in descending order

  38. Criteria of acceptability quality • Users have different assessment criteria for different content types • Movie – high quality required (HD quality) • News – audio quality and sync. • Music – audio quality • Animation – fewer requirement • Sport – higher quality needed when small objects appear • Users’ preference leads to different result on the same content type • Ex: sport and news

  39. Discussion and Conclusion • Users’ profile matters • The result is different from previous works • Exact required bitrate still depends on individual video, here only gives a estimated range • Platform dependency as well as video codecs • Fixed vs. adjustable service? • Prediction model and optimal delivery strategy

  40. The End • Thanks for your attention

More Related