[Ffmpeg-devel] [PATCH] increase max numbers of B frames
Thu Feb 23 11:10:45 CET 2006
On Wed, 2006-02-22 at 11:44 -0800, Loren Merritt wrote:
> On Wed, 22 Feb 2006, Erik Slagter wrote:
> > So this is what I did, I ran mencoder with this (x264enc opts) command
> > line with bframes = 0..5 (didn't try higher values, no interesting
> > results...)
> > -x264encopts "pass=$pass:turbo=$turbo:nofast_pskip:bframes=
> > $b:nr=500:bitrate=2000:vbv_maxrate=9000:vbv_bufsize=5000:scenecut=-1:threads=2"
> > pass = [0,1], turbo is 1 when pass is 0 and vv.
> > The result is that average qp goes down just a little bit when going
> > from 0 to 1 bframes, after that the difference really is insignificant.
> > Am I doing something wrong here?
> You mean, besides not telling it to print psnr?
OOOOPS! I really must learn to read AND remember messages.
> "nr" will greatly degrade psnr, whether or not it improves real quality,
> since psnr compares encoded output vs the non-denoised input. So if your
> video is noisy, instead run it though hqdn3d or something before
Interesting. Will do so.
BTW the source is from a DVB transport stream, remuxed to program stream
by projectx and mplex, codec is mpeg2 with a fairly high bitrate (BBC),
I guess around 6000-7000 kbps. It contains some zooming, panning, fog
effects and spotlights panning also shining partly into the camera at a
certain point. The source material only shows some blockyness at one
point, which I, of course, will not blame x264 for ;-)
I enabled denoising in an attempt to get better quality out of the
ffmpeg x264 interface and actually, it did (although minimally), and now
I understand why. But using mencoder I shouldn't need to use it all with
this material, got your point.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 2771 bytes
Desc: not available
More information about the ffmpeg-devel