Log in

View Full Version : Problem with SmartDecimate


BangoO
10th November 2004, 11:53
Hi,

To do some tests, I tried SmartDecimate on the intro of Gladiator's NTSC DVD.

The AVS script is the following (the field order is correctly set):

Loadplugin("C:\Program Files\AVIsynth 2.5\plugins\mpeg2dec3.dll")
loadCplugin("C:\Program Files\AVIsynth 2.5\plugins\smartdecimate.dll")
mpeg2source("D:\Gladiator\glad.d2v")
AssumeTFF()
SmartDecimate()

The result is good most of the movie, but on static scene, such as the intro where there is only text displayed, it looks terrible.

Here is a sample:
http://perso.wanadoo.fr/bangoo/badglad.avi

I tried setting the tel param to its max, it's better but still far from perfection (like when I use Telecide () and Decimate()).

If there is a way to do this perfectly, please tell me because I would really like to use SmartDecimate (it's so fast).

Thx ! ;)

PS: of course, I get this problem on al the sources I tested, not only on Gladiator.

scharfis_brain
10th November 2004, 19:21
If you have no orphaned fields in your video, use this solution:

mpeg2source("D:\Gladiator\glad.d2v")
AssumeTFF()
x=doubleweave()
SmartDecimate(24,60,bob=x,weave=x)

smartdecimate always falls back to dumb bob,
if not enough motion is present!
bob=weave=x is some kind of hard IVTC. non bobbing at all.

BangoO
11th November 2004, 00:45
Thx a lot, as usual... you are the man ;)

I tested it on the intro of Gladiator, it's almost as good as Telecide + Decimate.
Still, Telecide + Decimate still looks a bit better...
I tried on another material that caused some problem and it is perfect, so I guess that's it.

I guess you tested all that already, which way do you prefer to do a perfect IVTC ?
I like SmartDecimate because it's really fast, but I didn't play enough with it to really compare the quality to Telecide + Decimate...

One last thing... isn't there a way to tell SmartDecimate that we have a perfect 3:2 pattern in order to avoid the problem I had ?

scharfis_brain
11th November 2004, 06:33
One last thing... isn't there a way to tell SmartDecimate that we have a perfect 3:2 pattern in order to avoid the problem I had ?

then use avisynth's internal pulldown()-function!


for more weird material better use (especially, if too much combing is present):

x=tdeint(mode=1)
y=doubleweave()
smartdecimate(24,60,bob=x,weave=y)

BangoO
11th November 2004, 10:20
I can't use pulldown as some stuff have glitchs or drops, so instead of talking about "perfect 3:2 pattern", I should have said "perfect 3:2 pattern 99% of the time" :)

So I guess I'll stick with your solution using SmartDecimate, unless you have another way to do this ;)

BangoO
11th November 2004, 11:26
Ohoh, another problem now :D

I tried on a file that has lots of drops and glitches.

If I use:

AssumeBFF()
x=DoubleWeave()
SmartDecimate(24,60,bob=x,weave=x)

After a glitch, it looks like there is a problem of field order.

But if I use:

AssumeBFF()
SmartDecimate()

There is no problem...

BangoO
11th November 2004, 12:00
I think I finally found the holly grail :D

I read the topic about UnComb (trbarry's filter):
http://forum.doom9.org/showthread.php?s=&threadid=52333

He says:
Yep, UnComb should be able to handle most field shifts and order changes. (barring bugs, it's new alpha code)

I tried it, and it seems to solves all my glitch/drops/field order change !