[mpeg2] Re: [mpeg2]How to reduce MPEG-2 coding delay

["Rob Morris" <rmorris@cdlsystems.com>]

I'm not sure what 'significantly longer' than 1 frame means, but the nature
of
MPEG-2 requires a few frame latency. The reason is B frames cannot be
composed
until the following I frame is present. Consequently if the encoding is:

I BB P BB P BB I

then the second last B frame cannot be composed until the last I frame has
been received.
In fact the last I frame is sent in the stream before the B-frame.

Perhaps the latency is lower if only I & P frames are used, but I havent
tried it.

-----Original Message-----
From: mpeg2-bounce@linuxtv.org [mailto:mpeg2-bounce@linuxtv.org]On
Behalf Of Warren Young
Sent: Saturday, September 15, 2001 3:28 AM
To: LinuxTV MPEG-2 Mailing List
Subject: [mpeg2] Re: [mpeg2]How to reduce MPEG-2 coding delay


Ronan KERYELL wrote:
>
> Theorically speaking, if we use variable bit rate MPEG2, it should be
> possible to compress the frame vertically each time enough lines are
> received from the camera to build a span of 16x16 pixel blocks.

VBR doesn't enter into it.  Oh, sure, maybe VBR works that way for
I-frames, but for P and B frames, the whole notion of 16-scan-line
slices goes away.  Think motion compensation and all that.

The real issue is that the card isn't delaying the video by a single
frame, but by significantly longer than that.  However, I have also used
two other Kfir-based cards, the Hauppauge PVR and the Sigma DVR, and
both also show this video encoding delay.  It may be a fundamental
limitation of the encoding chip.
--
= Warren -- Video articles: http://www.cyberport.com/~tangent/video/
=
= ICBM Address: 36.8274040 N, 108.0204086 W, alt. 1714m