Wavelet-Based VBR Video Traffic Smoothing Dejian Ye, J. Cam Barker, Zixiang Xiong, and Wenwu Zhu...

Preview:

Citation preview

Wavelet-Based VBR Video Traffic Smoothing

Dejian Ye, J. Cam Barker, Zixiang Xiong, and Wenwu Zhu

IEEE Trans. Multimedia, Aug. 2004

Outline VBR video transmission Smoothing algorithms

MVBA (minimum variance) MCBA (minimum changes) PCRTT (piecewise constant-rate)

WTS (wavelet-based traffic smoothing) Smoothing at multiple resolution Low computational complexity

Performance evaluation

Typical video distribution scenario

VBR: single traffic flow

VBR: aggregated traffic flow

From server to client:

link 3

link 2

link 1

Buffer constraints

MVBA (minimum variance bandwidth allocation)

U

L

MCBA (minimum changes bandwidth allocation)

U

L

PCRTT (piecewise constant-rate transmission and transport)

Divide the video stream into segments to create a transmission schedule.

Each segment corresponds to one rate in the transmission schedule. (set the transmission rate to the average frame size for each segment)

Raise the transmission schedule to avoid client buffer underflow.

Force a rate change after every segment.

Seg 5Seg 4Seg 3Seg 2Seg 1

PCRTT

L

WTS PCRTT works well for both single

and aggregated flows. WTS is based on PCRTT.

Motivations: Smooth aggregated flows. (?) Client buffers can be heterogeneous.

(difficult to schedule transmission)

Possible schedules in WTS

Binary tree Tmax

Has enough buffer?

uj+1, 2k

uj+2, 4k+2

uj+2, 4k+3

Multiple-flow clients and movies

#flow: #rate changes

#flow: rate variation

#flow: the min time between two adjacent rate changes

#flow: peak rate

Buffer size: #rate changes

Buffer size: rate variance

Buffer size: the min #frames between two adjacent rate changes

Buffer size: peak rate

Recommended