interlaced material with convolution3d

[KoPGaMeR]

I'm encoding metropolis ntsc dvd..
Forced film in dvd2avi didn't work so I did a tmpegenc deinterlace job. Then I make a pseudo avi and encoded with nandub-sbc.
But it has some noise so I want to use Convolution3d filter and use avisynth.
So I make a avs with this avi..

LoadPlugin("G:\RiP\Convolution3DYV12.dll")
avisource("G:\output\metropolis full frame_tpr-vfapi.avi")
ConvertToYV12()
crop(4,14,708,452)
LanczosResize(608,320)
Convolution3d(0,4,4,4,4,2.8,0)

will it work ?
I can't change YV12 to anything else, I changed avisynth to ver 2.07 but didin't work, so I get back to 2.5..
do you have another suggestion with this pseudo avi ?

[Asmodian]

I encoded that disk and I was able to do a very nice inverse telecine with decomb and then used convolution3d which worked fine. I might have changed the telecide threshold/threshold2 settings though. This way you can avoid the color space conversions (to RGB right?!?) that happen when using a pseudo avi.
Except for the extra color space conversions your script should work fine. You may want to compare an encode using your script and one using BicubicResize(608,320,0,0.5) after convolutuion3d instead of lanczosresize before, I have noticed a significant drop in quantizers when changing from lanczos to neutral bicubic with very little/no change in video appearance (maybe just a bit softer but the drop in quantizers makes the final encode look sharper). I always use convolution3d before any resizers though.

[KoPGaMeR]

thanks for your reply..

I did a 2 pass encode with these settings, but it still have some noise; I think..
a little bit detailed from my other full-nandub rip though..
what setting did you use for Convolution3d ?

[Asmodian]

For that anime I think I used 0,6,8,6,8,3 but I am not sure. One can use stronger convolution3d settings if doing it before the resize.