[MEncoder-users] lavc or xvid

[Phil Ehrens]

Loren Merritt wrote:
> On Tue, 4 Sep 2007, RC wrote:
> > On Tue, 4 Sep 2007, Phil Ehrens wrote:
> >
> >> Eek! nr=300?! Alright, I admit it... I reported the same thing
> >> once (or twice). If the source genuinely has significant gaussian
> >> noise, nr=300 can perform miracles... HOWEVER, if you are starting
> >> with good clean source (like a well-made dvd), nr=10 to nr=30 is
> >> more realistic.
> >
> > Almost every DVD I've seen has significant noise, and nr=300 helps a
> > lot (and I've done several visual and PSNR tests to demonstrate that).
> 
> Are you really saying that nr=300 improved PSNR? It intentionally changes 
> the content, so it intentionally reduces PSNR if you compare input to 
> reconstructed output. Which is not to say that it's impossible: maybe you 
> save so many bits by removing the noise that when those bits are 
> reallocated to other regions of the movie they make up for the earlier 
> PSNR loss.
> 
> >> These numbers are for 2-pass encodes. If you are doing single-pass
> >> encodes, hqdn3d will behave much more sanely than nr. But if you
> >> are doing multi-pass, nr will be better on all live action. Anime
> >> seems to really love hqdn3d, however.
> 
> Interesting. The only difference between 1 and 2pass I can think of that 
> might cause that is that 2pass allocates bits as a function of bits (i.e. 
> magnitude of quantized DCT coefficients), while 1pass allocates bits as a 
> function of SSD. DCT coefficients have passed through nr, while SSD 
> hasn't. So scenes with little content but much noise will be allocated 
> lots of bits in 1pass nr, but not in 2pass nr nor 1pass hqdn3d.

Now, *that* is INTERESTING. If I am understanding this correctly
it means that in single pass mode, nr is doing little other than
lowpass filtering.

> > Unless it's extremely noisy material, I wouldn't use hqdn3d at all. With
> > default values, it causes blockiness, fades color, and leaves a very
> > noticable trail behind objects in motion.
> 
> hqdn3d doesn't have any block structure at all, so the only possible way 
> it could cause blocking is if your video was already blocky but hidden 
> behind all the noise, and hqdn3d just removed the noise.

That description seems perfectly consistent with my experience.
Furthermore, I don't see "trails" with hqdn3d except at VERY
high values. Certainly not at the default values, and not even
at 8:8:7:8, for example (with anime). When dealing with damaged
VHS source, trails are tolerable when the alternative is mush.

For live-action, I only use hqdn3d when the noise is SO intense that
tonal stepping is prefereable to turbid murkiness. nr is nearly always
preferable to hqdn3d for live-action.