Re: [MPlayer-dev-eng] key frame extractor / vo_jpeg use YV12 instead of RGB patch

[Andreas Krebs]

> On Thu, 15 Mar 2007, Andreas Krebs wrote:
> > this is how far I got, however this already created some interesting results, so vo_jpeg
> > now runs almost twice as fast on my PC, simply to the fact the it now directly uses
> > YV12 rather than converting to RGB and back, also  the image quality seems to have
> > improved a bit by this (at least the resulting jpegs look better to me - but maybe it's
> > only my imagination ;-) ).
> 
> It's much better to do it this way, sending YUV to libjpeg rather than converting
> to RGB and then having libjpeg convert back to YUV.
> 
> You are not doing it quite right though.  The YUV data in video formats is
> almost never full range, but has something called ccir r.601 excursions.
> JFIF standard jpegs on the other had have full-range yuv data.  So, you
> must convert one to the other if you want to create standard jpeg files.

I wasn't aware of this - thanks for letting me know. I guess this is the reason
why my images looked different than the images produced by the original vo_jpeg.

this means I would have to scale each Y value to full range, is there
already an algorithm available for this? or better can I define a different YUV
format, so that the data passed to vo_jpeg already correctly scaled?
the Cb Cr values are not scaled - is this correct?



_____________________________________________________________________
Der WEB.DE SmartSurfer hilft bis zu 70% Ihrer Onlinekosten zu sparen!
http://smartsurfer.web.de/?mc=100071&distributionid=000000000066