[FFmpeg-user] Lossless x264

[Rodney Baker]

On Tue, 17 Jan 2012 21:33:47 dE . wrote:
> On 01/17/12 03:01, Rodney Baker wrote:
> > On Tue, 17 Jan 2012 00:59:21 dE . wrote:
> >> [...]
> >> Or is it -qscale (for some codecs) and qp=<0 to 50>? I think these are
> >> better. These are the real quality scales right?
> > 
> > qscale = quantizer scale, not quality scale.
> > 
> > qp = p-frame quantization factor
> > qi = i=frame quantization factor
> > 
> > The higher the number, the higher the compression (and the lower the
> > quality). The q-factor has a direct bearing on the bitrate of the
> > output.
> 
> qp or qi... what's the end effect?
> 
> Thanks for the info!

I don't know  if you can separately specify qi or qp in ffmepg and/or x264 - I 
was referring generically to h264 video. To answer the question, though, since 
I frames are much larger than P frames (for which only changed information is 
transmitted), increasing the minimum quantisation factor for I frames relative 
to P frames can smooth out network bandwidth usage (obviously most useful when 
streaming video) and avoid the large peaks that can otherwise occur every I 
frame interval, without affecting picture quality (especially with motion) as 
much as high q factors on P frames.

Some hardware encoders (e.g. IndigoVision) allow separate control of Qp-
min/max and Qi-min/max. I don't think that's the case with ffmpeg and it 
probably isn't necessary for simple video conversion tasks. It might be a 
useful feature to have for FFserver but it would need to be implemented at the 
codec level i.e. the mpeg4 and/or h264 (libx264) codecs.

-- 
===================================================
Rodney Baker VK5ZTV
rodney.baker at iinet.net.au
===================================================