[XviD-devel] Multithreaded encode

suxen_drol xvid-devel@xvid.org
Fri, 20 Dec 2002 07:37:27 +1100


On Thu, 5 Dec 2002 21:21:15 +0100 (CET) Christoph Lampert <chl@math.uni-bonn.de> wrote:

> On Fri, 20 Dec 2002, suxen_drol wrote:
> 
> > 
> > On Thu, 5 Dec 2002 14:11:35 +0100 (CET) Christoph Lampert <chl@math.uni-bonn.de> wrote:
> > > 
> > > Can any of you tell me the latest results when using the  
> > > fixed quant 2 file from first pass as input for second pass? 
> 
> > its already implemented in the xvid vfw gui.
> > this method of encoding is generally used to improve speed. it is
> > quicker to decode mpeg-4 than decode/resize/filter mpeg-2, but there is
> > some minor visual quality loss.
> 
> Quality loss compared to what? Or, with other words, is this a _visible_
> loss or just lower PSNR? 

minor *visible* loss (softer image)..

> Because most people would consider a perfect reproduction of the quant 2
> file just as good as a perfect reproduction of the original, even though
> PSNR would differ. 

at quant 2 the dct coeffs are divided by 4. on the second pass, the loss
is compounded. ive done no psnr testing.

-- pete; life is like a box of ammo