AnimeIVTC() - All-in-one solution for interlacing and blending - v2.00 [2010-01-07]

[GibsonXXI]

The source is the R1 re-release of 3x3 Eyes. a very poor quality source. Lots of noise, dot crawl and other nasties abound in this one. The original was a VHS release, circa 1996 so i'm not expecting amazing quality from the encode. All i'm aiming for here is watchable quality without credits that jump up & down like a tree-monkey on acid.

:)

Gib.

[thetoof]

GibsonXXI, the numbers you specify should not be from an ivtced clip, but the original source at 30fps (after a quick glance at your numbers, it looks like this is the issue; I could be wrong)

Undead Sega, use mt_masktools-25.dll with avs 2.5.x

[GibsonXXI]

Should i set DGIndex to ignore pulldown flags? Other than this the only thing my script has from where i get my frame numbers is the dgdecode and colormatrix lines. Until i get my frame reference points, AnimeIVTC is commented out.

Gib.

[thetoof]

Weird. I'll be busy for the next days; if I ever forget to look into it, you're welcome to remind me!

[GibsonXXI]

I set DGIndex to ignore pulldown flags and created a new D2V based on that setting. Then the episode is shown in it's entirety. obviously once I decimate with AnimeIVTC the frame total is smaller, but all frames are present that i expect to be there.

Gib.

EDIT: Did some more investigation. It looks like Pioneer did all they could to cut costs when transferring this title from VHS to DVD. The source appeared to be telecined, but in fact it is field blended from 23.976fps to 29.976fps. The source has interlacing and ghosting only possible when field blending has been used. The original was NTSC so i'm going on the assumption they used this to create the 29.976fps on the DVD edition rather than a PAL -> NTSC transfer. Therefore i'm going to try mode=2 and am going to attempt to denoise as much as is possible, and convert the audio to match.

[Kaze_no_Hime]

Can someone help me with this http://www.mediafire.com/?kdehzdma5yy example. It's only the opening of Air TV, the anime I'm trying to encode. The source is hard telecined with 3:2 pattern - 3 progressive and 2 telecined(which are blended). You can see what the problem is from 01:15 min. to the end of the clip. I can't figure out how to get rid of those blended frames...
The script I used is supposed to be like this:
DGDecode_mpeg2source("E:\Demuxed DVDs\Air\air - 01.d2v", info=3)
ColorMatrix(hints=true, interlaced=true, threads=0)
AnimeIVTC(mode=3, killcomb=2, aa=4, dark=0.4, aapel=4, omode=1, mix=22, estart1=0, eend1=3488, pstart1=3489, pend1=6184, estart2=6185, eend2=39870, istart1=39871, iend1=42568, estart3=42569, eend3=43492)
I'm specifying both the start/end frames of the sections, because I'm currently using AnimeIVTC - v1[1].06.1

I have tried AnimeIVTC dev copy(of v2.00) with those settings:
DGDecode_mpeg2source("E:\Demuxed DVDs\Air\air - 01.d2v", info=3)
ColorMatrix(hints=true, interlaced=true, threads=0)
AnimeIVTC(mode=1, blend=true, credconv="mocomp", pel=4, nnedi2pel=true, killcomb=3, aa=4, dark=0.4, sharp=160, aapel=4, omode=1, e1=768, p1=3974, e2=19388, p2=41976, e3=43307)
I triedn "blend" instead of "mocomp" as well. But that didn't worked too.
Does anyone managed to remove blended frames with AnimeIVTC 2.00 at all?

[Kaze_no_Hime]

Then, a stupid question - how can I cut a .m2v file to give the original source?

[Mr VacBob]

The param for aatype is wrong in AnimeIVTC (as of v2.00); it's "int" instead of "string".

[Kaze_no_Hime]

Ok. I cuted the .mv2's opening in mpg file and changed the link in my previous post. So it's not mkv anymore.
I really would appreciate any help for this source.

[bizz & buzz]

When trying to load sharpaamcmod(), I get an error: aWrapSharp isn't loded! (AnimeIVTC 2.00.avsi, line 547).
I'm using aWrapSharp 2009.06.19
What's wrong?

[bizz & buzz]

yep, totally worked -- thank you for that :)

[ShocWave]

This may be a stupid question, but is it possible to run mode=1 without decimating? Like what you can do with ymc in yatta?

[Tima]

I have source with very weird type of telecine.

http://griffon.net.ru/~next/IVTC-foo.m2v

It's from original DVD "Chikyuu ga Ugoita Hi".

What is the most proper way to process it, and does AnimeIVTC already employs it?

Currently with mode=2 results are satisfactory, but maybe they could be even better :)

[osgZach]

You should post a source sample from the DVD's VOB stream then.

Use DGIndex to open your VOB and mark a short 10 second section with the bracket keys ( "[" and "]" )

Then use the "save project and demux video" option. Upload the resulting M2V file.

[Tima]

Fixed.

http://griffon.net.ru/~next/IVTC-foo.m2v

[Stephen R. Savage]

That's not a "very weird type of telecine." In fact, there is no telecine at all in there. This is where you should be looking. (http://forum.doom9.org/showthread.php?t=95924)

[Tima]

That's not a "very weird type of telecine." In fact, there is no telecine at all in there. This is where you should be looking. (http://forum.doom9.org/showthread.php?t=95924)

Currently, AnimeIVTC(mode=2) does FAR better job for me than trying to use Srestore, so I think I'm posting my sample in the right thread.

Also this source IS telecined, because of 29.97 fps and visible interlacing.

In addition there are some fields which are blended very uncommonly ('crossfading' into each other by Y-axis), so I suppose there should be a special treatment for them.

[Caroliano]

I'm having problems deinterlacing minipato. Samples and more information in this thread: http://forum.doom9.org/showthread.php?t=153095

Spline64resize(720, 240).SRestore(omode="pp3").Decimate().Spline64resize(720, 480) works resonably well, but I lose half of the horizontal resolution. AnimeIVTC(mode=1, blend=true) works well at the fishing scene and many others, but it fails at the robot scene, where AnimeIVTC(mode=1) works perfectly.

Isn't there anyway for animeivtc process with the blend=true and blend=false, and return the one that gives the best results for each frame?

[osgZach]

Currently, AnimeIVTC(mode=2) does FAR better job for me than trying to use Srestore, so I think I'm posting my sample in the right thread.

Also this source IS telecined, because of 29.97 fps and visible interlacing.

In addition there are some fields which are blended very uncommonly ('crossfading' into each other by Y-axis), so I suppose there should be a special treatment for them.

29.97 FPS + "interlacing" does not always mean Telecined.

Just from looking at the frames, each 5 frame block contains 4 with combing and 1 progressive frame. This is a pretty good hint it is not a telecine issue. Or at least that the likelihood of it being an odd Telecine patterns is probably less.

You do in fact, have a field/frame (w/e its called) blended source.

Do this is to see for yourself. The scene with the train passing will give you a clear example of blends.

*.AVS

mpeg2source("blahblah.d2v")
AssumeTFF().SeparateFields()


Play that back frame by frame in any section with motion and you will see a "ghosting" effect. This is what blending looks like.

You need to use SRestore, or another similar tool. I think AnimeIVTC may use it in some modes, but I can't recall.

[Tima]

29.97 FPS + "interlacing" does not always mean Telecined.

Just from looking at the frames, each 5 frame block contains 4 with combing and 1 progressive frame. This is a pretty good hint it is not a telecine issue. Or at least that the likelihood of it being an odd Telecine patterns is probably less.
If the original framerate is 24 fps, and we have 29.97 fps, then there are some fields duplicated (w or w/o another mess-ups, such as blending, but it's still telecine).


You do in fact, have a field/frame (w/e its called) blended source.
Correct, there's not ONLY telecine issue with this source. As I said, the issue with this source (at least for me) is MIX of somewhat exotic telecine AND exotic blending".

If we look at the title of this thread, it says 'All-in-one solution for interlacing and blending'. My case exactly fits into this title ;) so, no offense, but it's not a solution to redirect to srestore :)


You need to use SRestore, or another similar tool. I think AnimeIVTC may use it in some modes, but I can't recall.
As I said, mode=2 gives satisfactory results, but I still need thetoof to consider this sample and probably enhance algorhitms of AnimeIVTC, or to clarify that it already does its best :)

[Nightshiver]

And you need to be more open minded. If you do not want to take the advice of those more knowledgeable than you, you have no business being here. AnimeIVTC was created mainly for the solution to the many problems with ANIME, and it just so happens that it works WELL with other sources, but is not the one stop end all solution to everything. There are many cases were other tools, such as SRestore, do a better job than AnimeIVTC at accomplishing a certain task. If have searched any field-blend threads here, you will also realize that sometimes, or a lot of times, you just can't get rid of every single blend. That being said, if you really want some help, post your case in MoMonster's thread here (http://forum.doom9.org/showthread.php?t=95924)

[osgZach]

I'm pretty sure what you think you are seeing as interlacing, are the motion artifacts from the field blends. I can't find anything that looks like Telecine related combing after running it through TFM. All I see are horrid, horrid, monstrously horrid, field blends.

Frankly I'm more inclined to go with Stephen's original assessment, cause he knows way more than I do. I guarantee you that.

[GibsonXXI]

When trying to load sharpaamcmod(), I get an error: aWrapSharp isn't loded! (AnimeIVTC 2.00.avsi, line 547).
I'm using aWrapSharp 2009.06.19
What's wrong?

Make sure your using the *original* version of aWarpSharp not the rewrite. The rewrite uses the command syntax aWarpSharp2() instead of aWarpSharp() or vice versa.

I'm not sure which version AnimeIVTC uses as i have both.

[Caroliano]

Quick bugreport: if I use animeivtc(mode=1, blend=true, chrfix=true) I get 11.988fps output.

[manono]

To add to the chorus trying to set you straight:
If the original framerate is 24 fps, and we have 29.97 fps, then there are some fields duplicated (w or w/o another mess-ups, such as blending, but it's still telecine).
It was converted to 29.97fps not by adding in duplicate fields but by field blending. Saying something's been telecined implies it can be inverse telecined. It can't be IVTC'd; it can only be unblended.
Correct, there's not ONLY telecine issue with this source. As I said, the issue with this source (at least for me) is MIX of somewhat exotic telecine AND exotic blending".

To repeat, there's no telecine, exotic or otherwise.
...but it's not a solution to redirect to srestore
Why not, if SRestore does the job as well as it can be done, and more easily? It's up to thetoof whether or not he wants to have a look at your sample, but if you have something already proven to work with your source, why even bother him?

[YnK]

Hello everybody... I'm new here, and not experienced in video encoding, so I apologize if there's anything wrong with my post. (I also apologize if there's anything wrong with my English. Or if I'm posting in the wrong thread.)

I need some advice on dealing with an anime I'm trying to encode. It has at least 5 or 6 separate randomly placed sections that are 29.97fps progressive, and the rest of it alternates quite often between regular telecine pattern (3 frames progressive + 2 combed) and what I assume to be double hard telecine (almost every frame is combed). "Quite often" means "somewhere between every 20 and every 5000 frames, depending on the scene, total frame count being over 54000". Also, the progressive sections have 1-2 pixels of combing on the edges (badly placed fields?).

I originally tried to encode it via 2 pass VFR mode of TFM+TDecimate, and it does its normal job with progressive and telecined parts, but it ends up outputting weirdness in the DHT parts (quite a lot of combing is simply left in with no deinterlacing, no matter what parameters are used). I tested AnimeIVTC in the DHT mode on a small fragment of a DHT part, and it produced a clean, if slightly jerky, output (I believe TDecimate left DHT at 29.97fps, because it looked smoother in motion - but very, very messed up in appearance).

However, I'm not sure there's a way to make AnimeIVTC process the entire thing at once, since it doesn't seem to be able to accept so many fragments of the video (and with some more tests I couldn't get it to produce VFR at all for some reason... maybe I missed something in the manual?..).

Is there any way to fix this anime? Any help is appreciated.

[Nightshiver]

Post a sample.

[Tima]

It was converted to 29.97fps not by adding in duplicate fields but by field blending. Saying something's been telecined implies it can be inverse telecined. It can't be IVTC'd; it can only be unblended.

Hmm, I guess that was the issue with my definitions, which led to misunderstanding here. I thought that BLENDING is an operation, which doesn't change framerate, and that the process of duplicating fields and blending some adjacent frames of one field together is called TELECINE+BLEND (and the reverse procedure is called UNBLEND+IVTC). But well, your point with that implication is valid too, so I'm convinced.

So, everyone, sorry for misunderstanding.


Why not, if SRestore does the job as well as it can be done, and more easily? It's up to thetoof whether or not he wants to have a look at your sample, but if you have something already proven to work with your source, why even bother him?
Well, I'm trying to stay with my sample in this thread because of the description of AnimeIVTC, the first post of this thread :)
The thread was originally created after encountering multiple interlacing and blending related issues [...] . Its purpose was to gather as many samples as possible and to converge discussions about them to eventually create a function that could deal with all the issues in an optimal and easy to use fashion.
Even with the field-blending only, the sample still falls into that description, I suppose. So my intention was to wait for thetoof to consider it :)

Sorry, if such behaviuor annoys you guys here :)

If there'll be no reaction from thetoof, I'll of course consider all of the suggestions in this topic and either will try to master specialized Srestore solution, or will just simply happily use AnimeIVTC(mode=2) to get suboptimal but still acceptable result.

[YnK]

Post a sample.

Not exactly sure if that's what you mean, but here:

30fps sample 1 (http://www.mediafire.com/?lokxtwgwkot)

30fps sample 2 (http://www.mediafire.com/?izdmmjmuodj)

DHT sample 1 (http://www.mediafire.com/?tykhnyckzzq)

DHT sample 2 (http://www.mediafire.com/?m0ma5zlvjz0)

I hope the links work because I never used MediaFire before and might have screwed up something... Also I didn't upload regular telecine, because it's just that -- but I can cut out some random piece of it if you want to look.

Of course, there's much more of this, since, as I mentioned, the video changes quite often - this is pretty much where the problem lies.

(This anime also has everyone's favorite scrolling credits on telecined background... but that's the least of its issues, really.)

[Stephen R. Savage]

@YnK:

I took a look at "HT2-1-30FPSsample3.m2v" and it appears to be an interlaced pan. TFM recognizes all the frames as being interlaced and substitutes deinterlaced frames for them just fine. Likewise, TDecimate correctly identifies this as a VIDEO cycle. This could be deinterlaced to 60p for better results, but TDecimate does not support NTSC_double cycles, so you would have to do this by hand. Incidentally, this artificial horizontal pan is one of the few things that TGMC/MCBob can reverse with 100% accuracy.

I also took a look at "HT2-1-DHTsample2.m2v". I am not entirely certain, but I believe the background scrolls in a telecine pattern independent of the foreground object (the smoke thing). TFM also correctly recognizes this as being interlaced. The correct strategy is to decimate this, because the primary motion is the vertical pan, and jerkiness in it would visually eclipse the less significant smoke. Actually, they appear to be in phase after deinterlacing, so there is no problem. Some combing on the clouds passes through due to the incorrect metric that TFM employs for detecting interlacing (sharp = interlaced), but reducing MI to 40 or so catches that.

In summary, I don't see any problems with using TIVTC on this material. For better deinterlacing, EEDI2, nnedi2, or TGMC could be coupled, but aside from that, there is nothing unusual.

P.S: What cartoon is that?

[YnK]

I took a look at "HT2-1-30FPSsample3.m2v" and it appears to be an interlaced pan. TFM recognizes all the frames as being interlaced and substitutes deinterlaced frames for them just fine. Likewise, TDecimate correctly identifies this as a VIDEO cycle. This could be deinterlaced to 60p for better results, but TDecimate does not support NTSC_double cycles, so you would have to do this by hand. Incidentally, this artificial horizontal pan is one of the few things that TGMC/MCBob can reverse with 100% accuracy!

Will try. Thank you.

I also took a look at "HT2-1-DHTsample2.m2v". I am not entirely certain, but I believe the background scrolls in a telecine pattern independent of the foreground object (the smoke thing). TFM also correctly recognizes this as being interlaced. The correct strategy is to decimate this, because the primary motion is the vertical pan, and jerkiness in it would visually eclipse the less significant smoke. Actually, they appear to be in phase after deinterlacing, so there is no problem. Some combing on the clouds passes through due to the incorrect metric that TFM employs for detecting interlacing (sharp = interlaced), but reducing MI to 40 or so catches that.

Will try, although I'm honestly not very much familiar with how exactly TFM works... I never had to encode VFR prior to getting these DVDs... so maybe I'm just using wrong parameters. The first DHT sample in particular always looked bad in my encoding...

In summary, I don't see any problems with using TIVTC on this material. For better deinterlacing, EEDI2, nnedi2, or TGMC could be coupled, but aside from that, there is nothing unusual.

Will try... I have a lot of free time for experimenting, actually. :)

P.S: What cartoon is that?

Eh... well... It's one of the OAVs for Harukanaru Toki no Naka de series. I kind of expected it to be VFR, but I wasn't prepared for my first ever encounter with double-hard telecine. The samples are from the first episode out of three... the other two are better in that they don't seem to have DHT... instead, the second one has what I think is blended fields, although I might be mistaken. But I guess that's another topic entirely... (I don't remember if it has interlaced pans as well or not.)

Thank you very much for your advice. This video is quite messy, at least compared to other anime I encoded so far, so I'm not 100% sure if your suggestions will work on the rest of it. Will post again in case there are more problems, if nobody minds. :)

EDIT: The Results!

Well, the good part is that I fixed all interlaced pans. They're great at 60fps. The bad part is that DHT and other weirdly interlaced video is still ugly. In particular, somewhere within the rest of the episode there's a couple of scenes with interlaced effects (sparkle, moving transparent smoke or fog, etc.) over telecined background. Their combing is not much visible on the original video stream to begin with, but TFM's attempts to deinterlace them tend to screw up the background so it requires even more postprocessing. When I originally tried to encode VFR with TIVTC alone, simply running the two-pass script on the video, everything that wasn't progressive or 3:2 pattern looked bad (and TIVTC also did weird decimation in some places). In general, AnimeIVTC often does a better job with these cases, but sometimes it's just as ugly as with TIVTC.

Oh, and another issue I forgot about. There are scenes that seem to keep flickering at any chosen framerate. I'm not particularly sure what to do about it. Sample (http://www.mediafire.com/?z5lztuky3tm).

[egrimisu]

Hi, i have a source that have some blends and those are properly remobed usign mode=2, but some scenes are progresive and does not require ivtc, the v1 did a good job mentaning the progresive frames using timecodes file, but lacked in deblending(mode=2). how to solve this?

[Nightshiver]

Post a sample please.

[egrimisu]

sample:
http://www.megaupload.com/?d=B5SH5TTS

[egrimisu]

no solution?

[Nightshiver]

Sadly, no :( There wasn't anything I found that could completely fix it, not VFR, not various methods of blend removal. I suppose the last think you may want to do is post in momonster's thread "another restore function", as he would best know how to get you further along. Either that or get a better source is all I can tell you :(

[Undead Sega]

Hello thetoof, how are you? i hope all is well.

i hope u remember my problem and situation from a couple months ago, but i updated the masktools to masktools-v2.0a36, and changed to version 2.00 of your avsi, and followede the same script u suggested to me (animeivtc(mode=1, blend=true) i get an error message saying that there is no such function as 'average', and even when i place the Average .dll, it would work but i would still get those corrupted frames. Any ideas or something i might be doing wrong?

[egrimisu]

Sadly, no :( There wasn't anything I found that could completely fix it, not VFR, not various methods of blend removal. I suppose the last think you may want to do is post in momonster's thread "another restore function", as he would best know how to get you further along. Either that or get a better source is all I can tell you :(

ok but in animeivtc v1.xx mode=5 worked great for mode=1 i was hoping that the toof to import that function in v2 and make it to work with mode=2 to.

[Overdrive80]

Hi, i find a ussue. If i use this code:

DGDecode_mpeg2source("E:\DB_1\Opening\NF\ope.d2v", info=3)+\
DGDecode_mpeg2source("E:\DB_7\41\41.d2v", info=3)

Animeivtc(mode=4,chrfix=true,aa=0,omode=1)

I obtain this error:

http://img97.imageshack.us/img97/3671/captura2gv.jpg

Someone solution?

Thanks ^^

[ghscreenki]

Pardon me for this off-topic question:

Is it common for anime DVD makers to create a progressive source rather than interlace it which they normally do?

I have an anime DVD and AnimeIVTC failed to read it (mode=1) and upon closer inspection, it's progressive and I find it very odd because all anime DVDs I handled are interlaced.

[Mr VacBob]

Some of the more carefully produced DVDs (esp. some things by Gainax) are soft-telecined 24p. Haibane Renmei is actually 30p, but I've never seen another one like that.

[ghscreenki]

The funny thing is, its TV broadcast version and it runs at 24fps. It's very strange really, it's an anime with little animation going on and 80% of the time are all talking with nothing but their lips moving. It's understandable if there's CG animation involved, etc. Anime in question is Sasameki Koto. Well, anyway, thanks. :) At least there are DVDs like those.

[Leinad4Mind]

Hi there, I have here a sample: http://www.megaupload.com/?d=M5V5W6P0

This is a R2ES (Spanish). PAL. 25fps. Interlaced. Top Field Frame. (Yeah... the bad... the ugly... the real shit xD)

And I know that I could use an R2J or R1 without those problems. But I just want to know how to solve and learn a little bit more.

I'm not a complete noob, but... what is that:

http://img717.imageshack.us/img717/1145/problem1v.png
http://img530.imageshack.us/img530/1440/problem2.png

I'm using:

mpeg2source("C:\Users\Leinad4Mind\Desktop\SF.demuxed.d2v", idct=6, info=3)
ColorMatrix(hints=true, interlaced=true, threads=2)
AnimeIVTC(mode=2,bbob=4,aa=4,edimode="nnedi2",degrain=3,normconv=true,omode=1)


This fix the interlace and denoise the vid and do an anti-aliasing. But those problems in the screens I show. How can I remove them? And what is that? Some sort of ghost?!

Best Regards

EDIT: @thetoof: The "chrfix=true", is it already fixed?!

In attach I send the AnimeIVTC with some corrections spot by many and solved by thetoof in some new posts. :p


fixes as:
int "aablk", string "aatype"){

bool "ifade", float "sfthr", bool "sfshow"\

smartfademod(clip c,

default ( sfshow, false )

ifade = aa==2 ? false : ifade
sfthr = default ( sfthr, 0.4 )
sfshow = default ( sfshow, false )

[osgZach]

Looks like a frame/field blend issue.

Sometimes you can't fix everything, or at least not without getting a little more manually involved. If at all

[manono]

It's just standard Film to PAL field-blended stuff. You don't need to use AnimeIVTC to fix it. Based on the sample provided, just:

Yadif(Order=1,Mode=1)#or a better bobber
SRestore()

is enough.

[MatLz]

Leinad4Mind's sample.

Sometimes, a small variation like that:loadcplugin("yadif.dll")
mpeg2source("VTS_01_1.demuxed.d2v")
yadif(1,1)
convertfps(48000./1001.)
srestore(24000./1001.)does a better job.
It seems that's the case here: www.sendspace.com/file/h02bxq

[Great Dragon]

Hi guys.
I can't deal with this chroma bug.
http://j.imagehost.org/t/0122/chroma_bug.jpg (http://j.imagehost.org/view/0122/chroma_bug)
Sample (http://rapidshare.com/files/375361070/dt_sample.demuxed.m2v)

I'm trying mode=1 with chrfix=true but it doesn't work. I have a script error message.
Mode=2 can handle it but it ruins progressive frame quality.

[YnK]

Just want to ask one question because I'm not sure if I'm doing the right thing or not... Please forgive me if it's already been said anywhere.

Scene transition effects are often applied to the video after telecine. So if there is a telecined scene with movement fading into another telecined scene with movement, the pattern sometimes doesn't match, so you get three or four combed frames in a row. If there are three, TFM fixes it if you make it search for matches in all directions, but I couldn't get it to fix four. I got this thing fixed with AnimeIVTC mode 3, but I'm guessing this might not exactly be a correct way to do this (more like I just got tired of messing up with TFM). The video in question is pretty much entirely telecined, but these odd scene transitions happen rather frequently, so I would like to know how to handle them properly. Please?..

[Gavino]

I'm trying mode=1 with chrfix=true but it doesn't work. I have a script error message.
I believe there is a problem with chrfix=true, due to a clash of global variables between AnimeIVTC and SRestore. See this post.

[Great Dragon]

I believe there is a problem with chrfix=true, due to a clash of global variables between AnimeIVTC and SRestore. See this post.

Thanks for advise. I have a different error but I can menage it.
First, I've changed all lfr to alfr.
Second, I've change variable call ordering to remove "I don't know what "l01" means" error:

l10 = jmp ? l01 : l_v
l01 = jmp ? l12 : l_v
l12 = l_v
to

l12 = l_v
l01 = jmp ? l12 : l_v
l10 = jmp ? l01 : l_v

and


c10 = jmp ? c01 : c_v
c01 = jmp ? c12 : c_v
c12 = c_v
to

c12 = c_v
c01 = jmp ? c12 : c_v
c10 = jmp ? c01 : c_v

After all this changes chrfix seems to be working, but I have another chroma bug:
http://j.imagehost.org/t/0428/chroma_bug2.jpg (http://j.imagehost.org/view/0428/chroma_bug2)
The same bug I've got with TempGaussMC but I've menage to remove this frame from sequence by this decimation pattern selectevery(5,2,4).
But pattern=5 command in AnimeIVTC that is responsible for above decimation scheme can't do the same job for an none reason.

Here I've attached an already modified version AnimeIVTC.