another restore function

[manono]

I'm also seeing leftover blending and other 'dirty stuff'. I'm not sure you can entirely get rid of it. One thing though. By default SRestore gives you 23.976fps (for PAL sources) or 25fps (for NTSC sources). You don't need the 'TDecimate' at the end, and it will cause the framerate to become 19.181fps, which you don't want. Leave off that line.

And welcome to the forum!

[Vector13]

Thx for the welcome manono.

I put TDecimate() in at the end because i felt it removed more ghosting, but i didn't know it would affect framerate. Thx for that.

I'm hopeing there is something that can remove the blend leftovers and "dirty stuff", else this will have to do it i think.

[Overdrive80]

Hi, somebody help me with dvd source of this topic http://forum.doom9.org/showthread.php?t=161968

Thanks

[egrimisu]

hey guys,
i have a r2 japan anime dvd that have some blends i i try to remove it with srestore, ivtc was made with qtgmc because it has a more stable image than animeivtc.
how do i use srestore for deblend and bring the framerate back to 23.976? the script that i used and kept the framerate to 59.94 was:

qtgmc("fast")
srestore(frate=23.976,omode="pp3")

also in the srestore manuale dmode is not explained, what does dmode do?

Thanks.

[manono]

You bring it back to 23.976fps if the 'base' framerate really is 23.976fps. The field blended Japanese DVDs I've seen were made from PAL masters and the framerate was 24.975/25fps. You have to figure out the 'real' framerate first, before deciding what framerate to set SRestore for. You use SRestore's omode only if you have deinterlaced hard telecine, and not field blending. If you see interlacing, you don't want to use the omode.

For more advice a sample would be useful.

[egrimisu]

hi manono,
here is the sample http://www.megaupload.com/?d=CCB5L7J0

[manono]

You don't want to use omode with that one. I can't figure out what the framerate's supposed to be as there's not enough frames with movement every frame. I suppose you could make it 23.976fps, but to be sure you'd have to find a few seconds where every frame is unique and not a duplicate. Much of the sample is 8 or 12fps.

Oh, and I don't think you can unblend it completely as there are just too many blended fields.

Hi,
Regarding this sample I posted preiously in this thread, I found that

Bob()
Srestore()

was reasonably successful in restoring progressive 23.976fps, but if you have any recommendations for a different bob filter or tweaking the srestore parameters to get improved results I'd be interested to hear them.

[EuropeanMan]

i'm at wit's end with this source
here is a sample: http://www.sendspace.com/file/8pr90h

tried to restore to 23.976/25. with 23.976 i get tons of dupe frames, with 25 i get a dupe frame almost every 6/7 frames or so...
basic yadiff/srestore.

what else can you folks suggest? thanks in advance.

[manono]

It's definitely supposed to be 23.976fps. This works reasonably well:

Yadif(Mode=1)#or your favorite bobber
CDeblend()
BlendBob()
TDecimate(Mode=0,Cycle=15,CycleR=3)

[Didée]

Just tried a basic "bob(0,0.5).Srestore(frate=23.976,speed=-1)", and the result seems okay to me. I don't get anything like "tons of dup frames", the result is mostly smooth.

[EuropeanMan]

^ Didee, I had used yad/mod3 + sres/fr23 which gave me more dupes than sres/fr25...

Thanks manono...will check it out. Also will have to read on what speed=-1 would accomplish in the above setting.

edit: detail function for speed. thanks Didee

[EuropeanMan]

Ran both of the offered suggestions...thanks for that. I find that Didee's solution works better, especially where scene changes occur. However, manono's is QUITE fast, encoding wise Ran a lag w/ both against 2,001 frames and D's came out with one extra frame (1601 v 1600). OFF the top of your heads, 'why' could/did this happen? Would it be due to the bobbers used? or...

[EuropeanMan]

Off the top of my head, could something like this ALSO work?

tdeint(mode=1)
srestore()
tdecimate()
assumefps(23.976)

[manono]

Originally Posted by EuropeanMan

Off the top of my head, could something like this ALSO work?

No. Your script after the third line gives you 20fps, and then you speeded it up to 23.976fps. The result will be a jerky video that plays much too fast. Not to mention the out-of-synch audio.

When using standard SRestore, even with Speed=-1 added, I was getting duplicate frames and missing frames about 11 seconds in with the two men in the scene. That's why I suggested what I did, something that sometimes works when the field-blending is in an unusual pattern.

The one extra frame isn't anything to be concerned over. Even for the entire movie, at most there will be a single frame difference in the final framecount, nothing to worry about.

[kws53]

I have the Jimi Hendrix Live at Monterey documentary. And like virtually all documentaries that use old video or TV footage, I have a blend of scenes with telecined material and other scenes with simple interlaced.

For example after bobbing with Yadif, the historical scenes will be the AABBBCCDDD scenario of telecining and the new interivews will be ABCD.

Is there some "easy" way of handling this? Or do I need to do a scene by scene conversion?

Thanks in advance.

Kurt

[StainlessS]

FixBlendIVTC_v091a.avs supplied in R_Pack (18/11/2009) is broken:-

http://forum.doom9.org/showthread.php?t=95924

Seems that it has been left in editing/testing mode, does not work and leaves a metric
on the clip. v0.9b in the wiki works fine.

http://avisynth.org/mediawiki/FixBlendIVTC

EDIT: the problem is in lines 135,36

ScriptClip(source, " source.subtitle(string(rvc0)) ")
#ScriptClip(source, " counter==0 ? final : counter==1 ? source.trim(1,0) : source ")

Commenting out the first line and uncommenting the second will fix the problem.

[cobo]

Is there a way to set srestore or adapt it so that it recognizes and replaces blends that use a unique frame - that is where one of the frames making up the blended frame is neither the previous or next? That's where it seems to be failing for me no matter what settings I try.

[Zep]

Normally I just throw srestore or FixBlendIVTC at this stuff and it works well with defaults but not this time. It appears the video is 1p5i/b. All i have ever seen before is 3p2i/b and 4p2i/b. Could some else please try their hand at it? I made a clip of three different areas of the video that clearly show the crappy blends.

here is the link. it is a 11 megabyte file.

http://www.fileserve.com/file/s8yjuBD/blended.avi



thanks


.

[Didée]

That's a 25p->30p normconversion in convertFPS()-style. Use RestoreFPS for this. Also, see here.