[MEncoder-users] Noise reduction and interlacing: (hq)denoise3d vs. nr

[Ivo]

On Monday 02 July 2007 20:50, Pierre Catello wrote:
> 2007/7/2, Nico Sabbi <nicola_sabbi at fastwebnet.it>:
> > Pierre Catello wrote:
> > >>>-vf tfield=0,hqdn3d,tinterlace
> > >>
> > >>bad idea, better use -vf il=d,hqdn3d,il=i
> > >
> > >I trust you, but could you explain why il=d,...,il=i is a better way ?
> >
> > they mess with the framerate, and btw, try to see what happens :)
>
> tfields doubles the framerate but tinterlace should halves it.
> I just checked  with a large spatial filter such that its effect would be
> obvious :
> $ mencoder "RFID.mpg" -nosound -endpos 30 -v
> f tfields=0,denoise3d=10:0:0:0,tinterlace -ovc lavc -lavcopts
> vcodec=mpeg4:ildct:ilme:vqscale=2 -o "RFID.avi"
>
> The result is as expected, and the framerate is right.
>
>
> fil=d,hqdn3d,fil=i,softskip should be faster, but the man page says
>
> > that it could crash
>
> May be your "implementation" is the way to go and is faster, but I was
> simply trying to help the original poster and the point I was addressing
> was the effect of a spatial filter on interlaced content (in addition to
> a question regarding lavc internal denoising filter vs hqdn3d).
> My suggestion was may be a bad idea as you mentionned, I'm sorry and
> surely won't do that again...

IMHO tfields/tinterlace is prefered if you run a temporal denoiser. If you 
only (de)interleave as Nico suggested, the denoiser works with different 
data and will perform worse.

For example: (F=frame, f=field)

Input F1(f1,f2) F2(f3,f4) F3(f5,f6) etc...

Fictional denoiser uses current frame, and the next two frames.

If you use il=d and il=i:
    F1 is denoised by using f1<-->f3<-->f5 and f2<-->f4<-->f6
If you use tfields and tinterlace:
    F1 is denoised by using f1<-->f2<-->f3 and f2<-->f3<-->f4

--Ivo