[FFmpeg-devel] Scalable Video Encoding and FGS

[Erik van Zijst]

Hi folks,

It seems every once in a while someone comes along asking for Scalable 
Encoding support (FGS, SNR, SVC, etc), often because they want to do 
fancy video streaming. I guess this time it's my turn :-)

Some time ago I built a wide-area publish/subscribe platform targeted to 
live data streaming over unreliable, fluctuating networks (wifi, 
Internet, etc) that are inherently hostile to live streaming.

This platform works by first slicing up the stream in prioritized 
logical layers, send these over the network and let the network drop 
low-priority layers during congestion, instead of buffering. The result 
is a client that never has to wait or re-buffer even during heavy 
congestions. Works really great for real-time streaming financial data 
(high-volume stock quotes).

I think streaming video could also benefit from this platform/protocol 
when you apply some form of scalable encoding: have a small base-layer 
and multiple enhancement layers. I've looked at MPEG2 FGS which comes 
closest to my needs, but this has not been implemented in any serious 
codec AFAIK.

I'm now doing my own FGS implementation on top of ffmpeg (if this works 
out and people are interested, I will contribute it), but I'm finding it 
quite hard, so I'm wondering as to whether anyone out here has built 
some kind of scalable encoder, or would be willing to help.

Also, I've set up a company around this streaming platform, and I'd be 
open to financially support the ffmpeg team in implementing FGS, similar 
to what was done with the snow codec donations.

kind regards,
Erik van Zijst
-- 
File cabinet:
	A four drawer, manually activated trash compactor.