Open-source Real-Time interactive videostreaming solutions, CBR, low-bitrate

[Hombre]

Hello, buddies!

Year ago I tried to implement a part of h.264 standard (intra and inter) in Java. It was bad idea, though…hehe :)
One year passed… Now I am again in video compression game and I am searching for good open-source LIVE (webcam, low latency) streaming solutions for low bitrates (CBR, ~100Kb/s, CIF ) to customize it for non-commercial academic use (customizations for remote controlling of mobile robot) and may be also to contribute to it (conduct research on rate-control techniques, CBR). Do you know such?
In my pure Java webcam streaming implementation I have been achieving 2 seconds delay. (in constrast to required <500ms).

Is x264 still far away from current state of h.264 standard?

I know about VideoLan and Icecast.... Is Theora worth competition with h.264 in case of live videostreaming or h.264 outperforms Theora undoubtedly?

[unmei]

While x264 implements quite a lot of the standard, but it doesn't include the tools aimed at streaming such as switching frames because it focuses more on backup than streaming or broadcast.

Another thing, if i interpret your post correctly, you are going to encode the video on a mobile robot?

Depending on the CPU power it has available, using h.264 may not be feasible. I don't think there already are h.264 encoder chips, else they are probably very expensive. May be it is more reasonable to use a hardware h263 encoder or a very simple software encoder, not such a beast as a software h.264 encoder.

You probably should state what restrictions apply, if there are any.

[bond]

videolan and mp4live (from mpeg4ip) are live streaming solutions supporting avc (via x264) i know

x264 is obviously NOT "away from current state of h.264 standard", but spec compliant since the beginning
x264 also supports CBR

regarding theora vs. h264: qualitywise x264 beats theora qualitywise (that doesnt mean that every h264 encoder automatically beats theora qualitywise!)
still theora beats h264 "pricewise" as for h264 streaming you would need to pay licenses and for theora you dont

[Hombre]

Thanks for answers, buddies! Sorry, in my first post I ment NOT Live but Real-Time interactive videostreaming (<1sec). The mobile robot is merely a platform for a remote human-human interaction via webcams and microphones. But the mobile robot needs a motion controlling system integrated with a real-time communication system. BTW, the robot holds a P4 Windows XP on board for the controlling and communication system.

Actually I am a master candidate and my research thesis lies in a rate-distortion optimization and improving the H.264 coding standard for a real-time interaction.

still theora beats h264 "pricewise" as for h264 streaming you would need to pay licenses and for theora you dont

As I heard Theora is really "true free". Could you explain me the following situation regarding paying licenses? We want to use H.264 in a system which provides university students with remote access to a university laboratory. It seems like a non-commercial application. Do we need to pay licenses in a such case?

[foxyshadis]

http://www.mpegla.com/avc/AVC_TermsSummary.pdf

I think the last sentence of (b) (2) applies to you, ie, free until 2010 at least. If it grows out of a research project and becomes a commercial application with codecs on each end, you may be liable for (a) (1) codec fees, which is still free up to 100,000.

[bond]

As I heard Theora is really "true free". Could you explain me the following situation regarding paying licenses? We want to use H.264 in a system which provides university students with remote access to a university laboratory. It seems like a non-commercial application. Do we need to pay licenses in a such case?hm good question i cant really answer for sure. from a feeling i would say that you dont need to pay licenses for this as its non-commercial (users dont need to pay for it), might be below the "too-small-to-pay-licenses" limit anyways and as its propably used only internally in your uni noone from mpegla might ever realise it exists ;)

[gameplaya15143]

2 seconds delay.That is pretty darn good IMHO. To reduce that you would have to reduce the buffer size on the recieving end. With too small of a buffer, you will get buffer runderruns (which is very bad).

Can we see the source code for this project? It sounds interesting.

If you are looking for an alternative method, nsv sort-of supports h264 and aacPlus. I don't think the delay would be any less though.

[Hombre]

I have conducted an end-to-end delay testing experiment of VideoLan with the x264 encoder and improved caching settings but the best achieved delay is no less than ~1 second.

testing results presented on the following webpage
http://www.borovkov.com/videolan_realtime_streaming_delay_testing.htm



Can we see the source code for this project? It sounds interesting.
It is temporarly under redevelopment. I think I'll fix it soon and make it available.