Artifacts on Eva Platinum

[helix]

I'm having a bit of trouble getting rid of artifacts in various high motion parts of Eva. I've tried tweaking some parameters in TFM to no success. Here is the scrips and the examples. The TFM part of the script was actually generated by MeGUI, as I'm really not familiar with how to use it.

MPEG2source("d:\encoding\ep.1.d2v")
tfm(mode=1, pp=7).tdecimate(mode=1,hybrid=1)
msmooth(threshold=15, strength=4)
Import("C:avisynth plugins\avis\limitedsharpen.avs")Limitedsharpen(ss_x=1.5, ss_y=1.5, smode=3, strength=30)
Tweak(0.0, 1.15, -9.0, 1.13)

http://i2.tinypic.com/rlcn04.jpg
http://i2.tinypic.com/rlcmzk.jpg

[Zarxrax]

First of all, remove "hybrid=1" from tdecimate. Eva is not hybrid.

I think the articacts are probably caused by the postprocessing mode in TFM. Try changing pp=7 to pp=5.

If that doesnt help, then perhaps limitedsharpen is causing the artifacts.

Oh, and do you really need Tweak? The colors in the source should be perfect as to how they are intended to be viewed.

[berrinam]

The TFM part of the script was actually generated by MeGUI
First of all, remove "hybrid=1" from tdecimate. Eva is not hybrid.
Update to 0.2.3.2112 as soon as you can -- it should hopefully give you better recommendations with regards to your source.

[helix]

I tried removing hybrid before with no effect so I just left it in. And I tried running the script with only TFM and saw no difference. But since pp=7 is the default I'll try pp=5.

[foxyshadis]

That looks like motion ELA artifacts. Try using pp < 5, MI or mthresh, raise the mode, or just use Revgen's Method:


For TFF clips:

AssumeTFF()
Interp = SeparateFields().SelectEven().EEDI2(field=1)
Deinted=TDeint(order=1,field=1,edeint=Interp)
TFM(mode=6,order=1,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)


For BFF clips:

Interp = SeparateFields().SelectEven().EEDI2(field=0)
Deinted = TDeint(order=0,field=0,edeint=Interp)
TFM(mode=6,order=0,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)

[berrinam]

But since pp=7 is the default I'll try pp=5.Did you add the pp=7 parameter in yourself? Because MeGUI shouldn't add anything like that.


First of all, remove "hybrid=1" from tdecimate. Eva is not hybrid.Just to clear this up: hybrid=1 shouldn't actually have much in the way of negative impacts, should it? Since the source isn't hybrid, TDecimate shouldn't detect any sections as video, so there shouldn't be a problem.... right? When I selected the filters that MeGUI recommends, I tried to make them resilient to incorrect matches, so I just want to check if other people agree about that.

Sorry about the hijack.

[Zarxrax]

Just to clear this up: hybrid=1 shouldn't actually have much in the way of negative impacts, should it? Since the source isn't hybrid, TDecimate shouldn't detect any sections as video, so there shouldn't be a problem.... right? When I selected the filters that MeGUI recommends, I tried to make them resilient to incorrect matches, so I just want to check if other people agree about that.

Sorry about the hijack.

You are probably right. Of course, there is ALWAYS the potential that a filter like Tdecimate could detect a part incorrectly. However, I think the chance is rather small, and even if it did happen, it probably wouldn't hurt things too much.

Probably.

[foxyshadis]

You can do a few things to help keep TDec from identifying video is a pure film environment. You can raise vidthresh and/or concycletp to keep outliers from spoiling. I use default vidthresh and concycletp=3 on known hybrid personally, and vidthresh=2.5, concycletp=4 or 5 for suspected film.

And of course there are times when you just want to force film, when it's just some stupid credits or studio logo tacked onto the movie throwing it off.

[Mug Funky]

one thing you should know - there's 1 single black field in the intro sequence, that'll mess up most field matchers (though tdeint(0,1,tryweave=true) handles it well). also, in spite of a rather half-assed remaster, the film is quite shaky during episodes. the Eva Renewal set (japan only AFAIK) is quite a bit better.

ADV is just cashing in on a popular title... (watch for Cowboy Bebop remix btw - apparently remastered, but there's composite crawlies all over the series).

i'm not in a position to complain about cashing in though - my workplace repackages these exact releases for the australian market... (though we've got some kickarse exclusive artwork for cowboy remix that sets it apart from the geneon release).

[helix]

Well tweaking the settings in both TFM and tdecimate didn't really reduce the artifacts. I also tried using "Revgen's Method" with no difference. However, Forced film and using only TFM does seem to work. Is this a "proper" method? - I always feel like I'm getting something for nothing when it works out easily this way.

http://i1.tinypic.com/rm3foy.jpg (720x480 this time)

Edit:


ADV is just cashing in on a popular title... (watch for Cowboy Bebop remix btw - apparently remastered, but there's composite crawlies all over the series).


I've found it better to just expect things like this from ADV.

[berrinam]

When you say 'only TFM' do you mean you are xcluding the other filters, like msmooth and limitedsharpen? In that case, are you sure the artifacting was even coming from the deinterlacing?

Also, I'm not sure about this, but I thought that LimitedSharpen was intended for natural sources, and something like mfToon would be a better choice for cartoons.

[helix]

When you say 'only TFM' do you mean you are xcluding the other filters, like msmooth and limitedsharpen? In that case, are you sure the artifacting was even coming from the deinterlacing?

Also, I'm not sure about this, but I thought that LimitedSharpen was intended for natural sources, and something like mfToon would be a better choice for cartoons.

I meant only FTM and not Tdecimate as the framerate is already at 23.976. I'm pretty sure they are from deinterlacing, becuase I saw them with nothing but FTM and Tdecimate in the script.

I thought LimitedSharped was good for anime becuase of it's characteristic of not oversharpening that generates edge halos, which tend to be more noticeable on anime. I'll see how mfToon looks though; aren't mf's filters suppose to be... notoriously slow?

[foxyshadis]

No reason to use tfm if you use force film, unless you're worried about stray combs leaking through somehow; even then it won't effectively catch them. The d2v was mostly film anyway, I presume?

vmtoon is a faster than limitedsharpen, with the same supersampling, though it defaults to a much higher supersampling.

[helix]

Nope, looks all NTSC... At least thats what DGindex says (Honor Pulldown flags).

[Chainmax]

That looks like motion ELA artifacts. Try using pp < 5, MI or mthresh, raise the mode, or just use Revgen's Method:


For TFF clips:

AssumeTFF()
Interp = SeparateFields().SelectEven().EEDI2(field=1)
Deinted=TDeint(order=1,field=1,edeint=Interp)
TFM(mode=6,order=1,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)


For BFF clips:

Interp = SeparateFields().SelectEven().EEDI2(field=0)
Deinted = TDeint(order=0,field=0,edeint=Interp)
TFM(mode=6,order=0,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)



Is there any reason besides encoding speed to not use those all the time?

[Lil' Jer]

also, in spite of a rather half-assed remaster, the film is quite shaky during episodes. the Eva Renewal set (japan only AFAIK) is quite a bit better.

ADV is using the exact same masters as the Japanese did so that's pretty much impossible unless they were somehow able to add wobble during the encoding of their discs. The Japanese discs, though, are pretty close to being 100% progressive (there are a number of breaks though) and the encoding is slightly better, but the wobble is exactly the same.

[helix]

No reason to use tfm if you use force film, unless you're worried about stray combs leaking through somehow; even then it won't effectively catch them. The d2v was mostly film anyway, I presume?

vmtoon is a faster than limitedsharpen, with the same supersampling, though it defaults to a much higher supersampling.

Thanks for that advice about about mvtoon, it does seem to work better. But what do you recommend I do about the combing? Running TFM to get rid of them just produces jaggies. And it will probally look worse If I blend them.

[Mug Funky]

@ Lil' Jer:

i've not seen the renewal stuff, but i have it on good authority that they are a separate remaster. i could be wrong on that though. i'm sure the encoding is better though - i haven't yet seen an ADV encode which utilised pulldown detection in the encoder. i don't understand this because the encoder will encode all fields if it's in doubt, but i'm sure they have their reasons.

[Lil' Jer]

@ Lil' Jer:

i've not seen the renewal stuff, but i have it on good authority that they are a separate remaster.i could be wrong on that though.

Then they told you wrong. They are using the exact same masters for the Platinum release as those used in the Japanese Renewal set. To quote the original press release:

HOUSTON, April 20, 2004? ADV Films, the #1 producer-distributor of anime in the U.S., has set a July 27, 2004 release date for the first volume of Neon Genesis Evangelion ? Platinum Edition. Each episode of the enormously influential Evangelion television series has been officially revised and digitally remastered for this new release by original creators GAINAX to create beautiful, vibrant video and crisp, sparkling new English and Japanese 5.1 surround sound audio, creating the ultimate Evangelion DVD experience.

Not to mention when this first came up and people were wondering whether it was going to use the renewal masters or if it was a separate ADV remaster the rep dlw confirmed it on the AOD forums that it was the renewal masters they were using.

[pwnsweet]

I'm really sorry for reviving a 3 yr old thread, but I am currently looking at purchasing the 7 DVD Platinum Evangelion set but I am unsure on which version to get: PAL or NTSC. I have no intention of re-encoding the discs or anything, I just want the best version in terms of image quality.

Does anyone here know which one is better?

[Chiba]

I guess the NTSC will be better. In fact, the japanese dvds are better ^^ but need some buyers to comfirm

[Nightshiver]

Why would anyone even think about choosing crappy PAL over NTSC? That's a no-brainer right there.

[Adub]

Go with NTSC. I own that version, and have cleaned 'em up perfectly using AnimeIVTC just fine. PAL anime is usually field blended together to get 25fps. NTSC should give you a more accurate framerate when compared against the original film.

[MKVCrazy]

I also have some artifacts in fast motions (non-anime). I don't notice if he solved his problem or not.

Is there a tool to remove the artifacts beside increasing the bitrate?

[Nightshiver]

Start your own thread please and let this 3 year old one die.

[helix]

I can't recall if I actually fixed that problem, but I think not. Even after all the help I got here, and improvements over my initial results, my final encodes were still unacceptable during the high action sequences (If I remember correctly). In the end I got myself some encodes that, as I was told, had had some of the frames during those sequences manually edited in Photoshop to remove the artifacts. Now with h264, I would imagine that this sort of encode would be more do-able, Xvid was just a pain back then sometimes.

And if that guy is still deciding between PAL or NTSC, get the latter :)