[Live-devel] H264VideoStreamDiscreteFramer vs H264VideoStreamFramer vs other

[James Norris]

Hey all,

I know some parts of this have been covered in (many) other posts, I've
spent sometime reading through them so apologies if I've missed the crucial
post but believe this is unique.

I have a live camera source providing individual frames which I'm encoding
in the x264 library, I've setup an RTSP server using a custom
"OnDemandServerMediaSubsession", and a custom FramedSource (using
DeviceSource as a template).  To give an idea of how it chains together I've
provided "createNewStreamSource" and "createNewRTPSink" in the footer, but
AFAIK they follow the guidance in the FAQ.

I can watch this stream using either the Discrete or normal framer classes
in VLC over the RTSP server when the x264 library provides annex-b
supporting NAL units (with the 00000001 start code) AND I do not try to
separate the NALs into individual calls to "doGetNextFrame()" - each frame
is provided as a single concatenated block of data where several NALs could
be within.  However there are problems with using both the discrete and
standard framers:

- If I use the H264VideoStreamFramer class, the fMaxSize variable counts
down till a frame is truncated, this truncation is visible in VLC as a
broken frame.  This problem is similar to
http://lists.live555.com/pipermail/live-devel/2010-July/012357.html where it
is advised to use a Discrete framer.

- If I use the H264VideoStreamDiscreteFramer class, I get the warnings about
a startcode being present, looking at the code this means that saveCopyOfSPS
and saveCopyOfPPS are never called.  It does play in VLC, I'm just concerned
about the implications of never using these functions?  If I remove the
startcode (just provide the remaining data block), VLC won't display
anything & in the messages it says "waiting for SPS/PPS", this is true
whether or not I split the NALs into individual "doGetNextFrame()" calls,
but in this case live555 seems happy and doesn't output any warnings.

- I've seen hints at writing your own framer class, but it's unclear why &
what I need to achieve in doing so?

Thanks in advance & do appreciate all help,
James


--

FramedSource* FramedServerMediaSubsession::createNewStreamSource(unsigned
clientSessionId, unsigned& estBitrate)
{
estBitrate = 500;

// Create the video source:
DeviceParameters p;
RTPFrameLoader* frameSource = RTPFrameLoader::createNew( envir(), p );

// encoder outputs to RTPFrameLoader
encoder = new H264Encoder ( frameSource, 640, 480, 3 );

Camera *cam = CameraFactory::getInstance()->getCamera( CameraFactory::FAKE
);

// the encoder listens in for raw camera frames
cam->registerFrameListener ( encoder );

encoder->go();

// Create a framer for the Video Elementary Stream:
return H264VideoStreamDiscreteFramer::createNew ( envir(), frameSource );
}

RTPSink* FramedServerMediaSubsession::createNewRTPSink ( Groupsock*
rtpGroupsock, unsigned char rtpPayloadTypeIfDynamic, FramedSource*
inputSource )
{
return H264VideoRTPSink::createNew(envir(), rtpGroupsock,
rtpPayloadTypeIfDynamic);
}
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.live555.com/pipermail/live-devel/attachments/20110607/761d84f0/attachment.html>