DVD of TV source terrible backup with Megui

[markrb]

Yes me again.
I have been trying to backup NTSC Lois and Clark Superman, but the output comes out pretty bad using Megui's suggestions for deinterlacing. TIVTC + TDeint(EDI) slow.
Still noticable interlacing.

Source declared Hybrid film/progressive.
Goes on to list the number of Progressive and Interlaced sections.

I am going to try it again with Decomb, but was wondering if there are better or other things to try.

Leaving it interlaced was also a no go.

Thanks,
Mark

[berrinam]

It's very surprising that it is still interlaced after a deinterlacer (actually, a field matcher) processes it. Perhaps TFM is making some wrong matches. Could you
Post some screenshots of the interlacing. Sometimes it stuffs up if there is low contrast
Post the generated script. I just put them together from what I read in various manuals, but maybe if you post it foxyshadis can fix my mistakes again and tell you if there's something wrong with it.
Post the number of progressive/film sections. This is just out of interest, because it may be entirely film after all, and the analysis was just slightly off.

Also, TIVTC should be better than Decomb, so if decomb fixes it, then there is probably something wrong with TIVTC's settings.

[ChronoCross]

I really dislike what EDDI2 does. the tfm().tdecimate() always has done a better job even at defaults on all my tests. Try the othere method suggestions in the drop down menu. It may give you better results.

[markrb]

Unfortunetly I erased the original file.
Basically it looked like horizontal lines like they were a frame apart in the same frame.
I could redo the avs creation later and post the script.

I did a quick play the decomb avs in media player and although very slow it did look better. I will know more when the encode is done.

Mark

[berrinam]

@ChronoCross: Maybe you are right, but conceptually, I don't see why. TFM().TDecimate() tries to match fields and if it fails, interpolates using something like cubic interpolation, (I think). TIVTC + TDeint(EDI) tries to match fields with TFM, and if that fails, requests deinterlaced frames from TDeint, which interpolates with the more advanced Edge Directed Interpolation. Maybe there's something wrong with EDI itself, but other than that, there shouldn't be any problems, really, should there?

[ChronoCross]

I think it has to be a problem with EDDI. Unless something isn't working in Tdeint the way you think it is. I wonder if trictical would know more.

[foxyshadis]

EEDI2 does cause some artifacts, and sometimes isn't as sharp as a kernel deint (esp on real-life footage), which is why kernel is default in tdeint. I'm not sure if that's what you mean. (Thus tritical's working on his new version... sadly he's sidelined by work now.) It might be that this is a noisy source that eedi just bombs on, but my suspicion is that it was resized and printed to dvd interlaced, which looks terrible when you try to deinterlace it. I used this after normal deint/ivtc for one:

tdeint(mode=2,full=false,cthresh=15,edeint=bob(),map=0)

Bob is so dumb that it works best on the dumb conversions! :p I may be way off though.

[markrb]

So far tfm().tdecimate() has shown the best results.

I will try bobomatic (tdeint(mode=2,full=false,cthresh=15,edeint=bob(),map=0)
tomorrow night.
It is just Lois and Clark so yeah pretty dumb. Must be the right choice then?

However the entire DVD is not interlaced only portions so will this screw up the non-interlaced sections?

I haven't been into something like this in a long time. Feels good to feel dumb again and realize there is
so much more I need to know. To bad I am working now. When I started with mpeg I was out of work
for almost 2 years and had plenty of time. Now not so much, but I can eat. So there is that.

Thanks,
Mark

[foxyshadis]

It's meant to come after the ivtc, and I wouldn't recommend it at all unless it actually is slightly vertically resized, it was just an off-the-cuff guess without seeing anything. (you can tell because it looks normally interlaced, but if you separate fields each appears to be composed of a little of both fields, in a wavy pattern. it's a kind of neat effect, try it sometime, resize 480 to 486 or so.)