[Live-devel] duplicating vop start sequence frames in openRTSP client?

Ted Morris tmorris at umn.edu
Fri Oct 14 07:01:50 PDT 2011


Greetings,

We use a Sony HCM-CH240 3mpixel camera to capture our data and MPEG4 
stream via RTSP to a  Linux box (Ubuntu) running openRTSP, -i to pipe 
output in *.avi format.  The problem is that the camera sends vop (video 
object plane) start sequence frame type-- I think in the same RTP packet 
payload as an I-frame (I-VOP) type (defining the start of a Group Of 
Pictures) We were sent proprietary white papers from Sony describing the 
theory of operation of this camera (for developers). Generally a bit 
stream of a  visual object sequence (e.g., Group Of Pictures) is 
terminated with another frame containing a VOP termination code;
but this termination code frame is not added in the case of this camera 
(I do not have enough experience with commercial ip cams to understand 
if this is the norm or not...)

The camera consistently writes out a vop start of sequence frame in this 
manner for *every* GOP sequence. It seems like openRTSP client keeps 
*BOTH* the 'empty' start sequence vop frame, *AND*, the I-FRAME, and 
actually assigns a time stamp to both??   Then, then for the next frame 
received in the payload, the program sees that the  timestamp matches 
the time stamp of  the I-FRAME assigned previously  and it is  thrown 
out to preserve the proper frame rate
Could this be possible??

The result: constant pause-and-skip motions of image sequence. I have 
now noticed this with other programs too  (for example VLC, etc.).  This 
has inspired us to try different tricks  with mencoder but none of them 
will remove  the duplicate, vop start-sequence frame--even though 
mencoder recognizes the problem!  I tried hard-setting framerate and 
frame size and it does nothing.

Has anyone encountered a similar problem, and if so, could you share how 
you addressed it?
I'm quite a newbie with utilizing openRTSP, and ip cameras for this 
purpose.

thanks to all in advance,

t


More information about the live-devel mailing list