Content Aware
video-quality-Fair (CAF) streaming
Ying Li, Zhu Li, Mung Chiang and A. Robert Calderbank
Y. Li, M. Chiang, and A. R. Calderbank
are with Electrical Engineering Department,
Z. Li is with Multimedia Research Lab, Motorola Labs,
Video clip consists of a sequence of frames with different importance.
When the network is congested, if the frame with high importance is dropped,
the perceptual video quality can degrade a lot, but if the frame with low
importance is dropped, the video quality degrade may be very small as perceived
by viewers. So the frames with high
importance should have higher guarantee to get through the network when the
network is congested.
In addition, different video clips may have different importance
distribution of frames. For instance, a video clip (User 1) with a lot of
motion may have many frames with higher importance than a video clip (User 2) with
small amount of motion. When they compete for a bandwidth, by the criterion of
bandwidth fair sharing, each user gets half of the bandwidth. But if the
network is congested and some frames of the video should be dropped,
perceptually User 1 is less happy than User 2 about the dropping because his
content is sensitive to the loss, but User 2 can be happy even if a lot of
frames are dropped because the content is not sensitive to the loss. To make
both users happy perceptually, User 2 drops more frames to help the frames of
User 1 to get through, such that both users can experience fair playback
quality.
CAF approach is fast and practical with content-aware cooperation, it does not violate
net neutrality since it is business-neutral yet technology-enabling.
Experimental results show that the proposed approach yields better quality of service
when the network is congested compared with the traditional approach of
content-agnostic networking.
Related article: http://www.technologyreview.com/Biztech/20394/