decomb questions

[jhr1948]

hi there
i'm a total noob so be gentle. i'm converting a pal dvd to ntsc dvd.
for the 2nd time the first time i used xesdeeni's older script (this pal dvd isn't progressive) w/decent results.
LoadPlugin("MPEG2DEC.dll")
LoadPlugin("SmoothDeinterlacer.dll")
MPEG2Source("DRIVE:\PATH\VTS_xx_x.d2v")
SmoothDeinterlace(doublerate=true)
LanczosResize(720,480)
ChangeFPS(59.94) # or ConvertFPS(59.94)
SeparateFields()
SelectEvery(4,1,2)
Weave()
ConvertToRGB()
slightly choppy to the picky eye, but not too bad and didn't expect an 1:1 copy (as far as quality). i saw this new pulldown, by neuron2, and wanted to try just for expirience. so far looked alright until i wrote my script (98% of which i got off fulcilives). he noted that if it is not pure interlace not to use FieldDeinterlace() instead use Telecide. this is where i realize i'm a total noob. so i wrote the script:
LoadPlugin("C:\Documents and Settings\All Users\Documents\Downloads\DGDecode.dll")
LoadPlugin("C:\Documents and Settings\All Users\Documents\Downloads\Decomb521.dll")
Mpeg2Source("7.d2v")
Telecide(order=0,guide=2)
LanczosResize(720,480)
ConvertToYUY2(interlaced=true)

is that wrong? after encoding i realized there was "vthresh" and "post" i could have added. did i not deinterlace? the video looks alright, but i want to make sure i did it all right.

i added more(after further reading the tutorial), but haven't encoded:
Telecide(order=0,guide=2,post=2,vthresh=30)

any help would be awesome
thanks

[Guest]

We'd need to see the source video to know whether you should use Telecide instead of simple deinterlacing. If the source is pure interlaced, you'd want just a deinterlacer. If it is progressive but with one-field phase shifts (so that you see combed frames in some sections), then Telecide is appropriate.

The ConvertToYUY2 looks doubtful if you are going back to DVD.

[Mug Funky]

i'm guessing the converttoyuy2() is there in case he wants to do a blend-conversion. i do the same thing :)

i wish i had the programming skillz to mod convertfps for yv12 (and give it scenechange detection as well - that'd be very cool indeed...).

[edit]

looking closer, it's more likely it's for feeding CCE with

[jhr1948]

thanks for the replies
i am feeding this into cce

neuron, would it be possible for me to send a clip to you to help me determine what the best way for interlacing. as i stated, i am still very much a noob.

there are both interlaced and progressive frames, that i do know.

thanks for the replies guys

[Guest]

You can put an unprocessed source clip on my ftp site (see sig below) and then notify me here, telling me the file name.

[jhr1948]

hey neuron2 thanks a lot. clip is called "ripping friends clip.mpv"

as i said before, i've done the changefps (xesdeeni's script for interlaced PAL dvd to NTSC) before and results were alright. however, i really wanted to try your new pulldown to compare the results. i just don't know what the best way to deinterlace. one thing i noticed is that when i had order=1 which is TFF i wasn't getting that many in frames, until i changed it to order=0 then it seemed on. dvd2avi sees it as TFF. as you can see i'm very much a noob. i've never gone this far, which is elementary things to most you guys.

thank you so much for the help

[Guest]

It's field blended video from an NTSC->PAL conversion. Do a search for "blended fields" to find out what your options are. Fortunately, I've personally never needed to work with blended field garbage; therefore, I have no recommendations. I would just deinterlace it, although others here think deblending is a viable option.

[jhr1948]

well i guess i have to do more homework

however, you do suggest FieldDeinterlace()

would this be a correct script if field deinterlace is what i go with:

LoadPlugin("C:\Documents and Settings\All Users\Documents\Downloads\DGDecode.dll")
LoadPlugin("C:\Documents and Settings\All Users\Documents\Downloads\Decomb521.dll")
Mpeg2Source("7.d2v")
FieldDeinterlace()
LanczosResize(720,480)
ConvertToYUY2(interlaced=true)

thanks for the help so far

[Guest]

Your script is OK. You might want to try some other deinterlacers to see if you get better quality, FieldDeinterlace() is intended as a postprocessor for progressive video. You might look at TDeint, (Leak)KernelDeint, TomsMoComp, and even my SmartDeinterlacer for VirtualDub (with the EDI option checked).

[MOmonster]

I prefer using telecide or tfm with postprocessing (tfm has a better postprocessor, but I like telecide more as matcher).
Functions, made by Didee and Sharfis_Brain could be also useful. For example the heavy slow restore24. Or also useful, the Leakkerneldeint with unblend and FDecimate. You have the choice :D

@neuron2
Are you still thinking about a further development of your telecide?
I think about a special mode for your telecide, that should give more smooth motion. Till now your telecide decide to choose c or n (order=1) depending on there combing levels, also if there is no possible match. What, if your telecide would decide to choose c or n depending on the last match.
Let me explain it short.
1.If we have a n-match or a c-match but with the possibility of a n-match, in the next frame we will first take a look for a c-match. If there is no possible c-match, we just take the n-match (without need to test the combing level) and make it for the next frame the same way. If we get a c-match, we do it like now and test also the n-match and compare the combing levels. If both is possible, we have the same frame but the match with less combing has the better quality.
2.If we have a c-match without the possibility of a n-match, in the next frame we will first take a look for a n-match. If there is no possible n-match, we just take the c-match (without need to test the combing level) and make it for the next frame the same way till we get a p-match. Then one more time we just compare the combing levels.
Is it understandable what I said (sometimes I write cruel sentences)?

So, for what should this mode be useful.
First of all, it should be faster in a way, because in this mode we don´t have a look to all matches (we choose one, because the other is not possible).
And the second more important point. If we have frames without matches sometimes I get something like this: ccccpcccpp...
But for the postprocessor a constant match use should result in smoother motion.
Please tell me, if you think this is useless for you, because you don´t have such sources.

[Guest]

That falls into what I call pattern guidance. Your rules sound very arbitrary. You'd have to justify them. For example, I justify the cccnn rule simply through the existence of 3:2 pulldown. To what do you appeal to justify these rules?

[MOmonster]

Most of my souces don´t follow a regular pattern. So to justify my rules to a guidance pattern, want help me a lot. There are many blended fields in my source. With this mode I just try to save all fields for the postprocessor.
Here is an example, where this mode could help to save field b.

source: a a b c d
a a/b b/c d d

telecide: a a c d d
(example) a a/b b/c d d

match: c c n n c

new mode: a b c d d
a a/b b/c d d

match: c n n n c

[Guest]

Sorry but I'm just not interested in making hacks to try to address pathological video, i.e., blended field video, especially when the proposed rules cannot be justified theoretically. How do you know that your rules would still improve things if the blending pattern was a little different?

[MOmonster]

That is what I thought :sly:
Ok this won´t help much. It all depends on the source.
I´ll try to find other ways.
Thanks for your replies, neuron2.