PAL sucks batdicks?

[Autraya]

Source: PAL DVD



My Script:

mpeg2source("F:\Utaw\D1\UtawD1test.d2v",cpu=4,iPP=true,idct=5)
#Fielddeinterlace()
Deblock()
tomsmocomp(1,5,1)
undot()
TIsophote()
deen("a2d,1")
tweak(sat=1.1)
vmToon(strength=24)
Levels(16, 1, 255, 0, 255, coring=true)
assumeFPS(29.970)
#ConvertToyv12()
LanczosResize(1024,576)

So after talking to coderjoe I tried a few settings in my import line and back with DGIndex (having problems jumping between versions of the amvapp but that's another long story), it improved somewhat, yes that does mean what you think it was worse so a little halp anyone?

it should look like this and 95% of the scenes do... the rest resemble that abomination above. With all my filters disabled it still looks fuggin terrible, so anyone know how to kill the artefacts other than not using PAL

[BasharOfTheAges]

Is it just the intro/ending credits that are different than the actual episode footage?

[mirkosp]

I tried a little script on the image you posted:

Code: Select all

vaguedenoiser()
FRFun7(lambda=1.8)
crop(2,2,-2,-2)
lanczosresize(800,450)
blur(0.8)
Deen("w3d",2,4,6)

the outcome:


Still not awesome, but looks better, I think...

[Autraya]

Is it just the intro/ending credits that are different than the actual episode footage?

it's throughout the anime, it was just easier to find the intro scenes when looking through 26 espisodes.

The srcipt is now looking like this:

mpeg2source("F:\Utaw\D1\UtawD1test.d2v",cpu=4,iPP=true,idct=5)
Deblock()
tomsmocomp(1,5,1)
undot()
TIsophote()
deen("a2d,1") <---- this removes the vague noise much better and it's a new series and quite clean so vaguedenoiser was a bit over the top
FRFun7(lambda=1.8) <--- what does this do? Never heard of it
blur(0.8)
Deen("w3d",2,4,6)
tweak(sat=1.1)
vmToon(strength=24)
LimitedSharpenFaster(strength=100) <--- needed some sharpening as it was pretty blurry still
Levels(16, 1, 255, 0, 255, coring=true)
assumeFPS(29.970) <---- this makes it easier to send to american cons
#ConvertToyv12()
Crop(4,4,-4,-4) <----- I usually crop once I'm finished editing unless require for the resize, but I added it anyway -_-
LanczosResize(1024,576) <----- keeping the proper resolution for PAL 16:9 not using your suggestion sorry[/img]

now my only concern is the over filtering, so if there are any alternatives please keep the ideas comming:

[mirkosp]

frfun7 is a very cool heavy denoiser. It's just spatial, so I used the vaguedenoiser to have a temporal there.
Now, about the script...
Uhm, why using the undot? You are using the deen for the light noise and you are removing the heavy noise with the frfun7, so no need to have too much denoise.
Also, if it's too blurry, just remove the blur(0.8). I threw it there as the frfun7 kinda ruined the borders in the image you posted, and it wasn't a matter of sharpening to solve that...
And plus, as you already use vmtoon, why using LimitedSharpenFaster too? Why not just incrementing the strength in the vmtoon? But then again, if you remove the blur(0.8), you won't probably need more sharpening either.
Also, instead of the deblock, try to see if the cpu=6 instead of cpu=4 works. It should remove the blocking if there isn't too much, but it doesn't blur. Deblock instead tends to blur a bit and removes details.
About the resize... egh, my bad, it was 800x450 just for the image purpose. XD
But I'd rather go with the 848x480 resolution... 1024x576 might be too big.
Try to see if my suggestions work... I'm not very good in filtering, so I might just have said bullshit...

[NekoMayumi]

http://www.flickr.com/photo_zoom.gne?id ... 125&size=o

My try at that pic. More conclusive ideas could be made with an actual video sample. (cut m2v from DGIndex or cut lossless avi from an only IVTCed source)

ImageReader("C:\PALisass.png").ConverttoYV12()
TBilateral(diameterL=9, diameterC=9, sDevL=3.8, sDevC=3.4, iDevL=12.0, iDevC=12.0, kernS=2, kernI=5)
FFT3DFilter(sigma=3.5, plane=3, bt=1)
DeHalo_Alpha(rx=1.0, ry=1.0)
Crop(4, 4, -4, -4).
BicubicResize(1024,576)

[BasharOfTheAges]

Is it just the intro/ending credits that are different than the actual episode footage?

it's throughout the anime, it was just easier to find the intro scenes when looking through 26 espisodes.

So, are the problems only noticeable in some scenes but always there, or are there significant noticeable differences that only occur in certain spaces? If it's the later, you might need to cut the episodes up and filter the segments differently.

[Autraya]

Is it just the intro/ending credits that are different than the actual episode footage?

it's throughout the anime, it was just easier to find the intro scenes when looking through 26 espisodes.

So, are the problems only noticeable in some scenes but always there, or are there significant noticeable differences that only occur in certain spaces? If it's the later, you might need to cut the episodes up and filter the segments differently.

only in some scenes and always there (how many scenes though I'm uncertain as I haven't looked at every scene obviously) looks like 2 sepperate scripts are going to have to be the way to go and I'll just encode as I need a scene

frfun7 is a very cool heavy denoiser. It's just spatial, so I used the vaguedenoiser to have a temporal there.
Now, about the script...
Uhm, why using the undot? You are using the deen for the light noise and you are removing the heavy noise with the frfun7, so no need to have too much denoise.

good point it is still need in the original script but not in the one to get rid of the artefacts *removes*

Also, if it's too blurry, just remove the blur(0.8). I threw it there as the frfun7 kinda ruined the borders in the image you posted, and it wasn't a matter of sharpening to solve that...
And plus, as you already use vmtoon, why using LimitedSharpenFaster too? Why not just incrementing the strength in the vmtoon? But then again, if you remove the blur(0.8), you won't probably need more sharpening either.

I was using the limited sharpen to compensate for the bluriness as the sharpening in vmtoon wasn't enough, i just tried it without the blur and the splothches are much more noticeable. It's already over filtering so increases to vmtoon are horrifying.

Also, instead of the deblock, try to see if the cpu=6 instead of cpu=4 works. It should remove the blocking if there isn't too much, but it doesn't blur. Deblock instead tends to blur a bit and removes details.

already tried that when talking to coderjoe i found the deblock was more effective
About the resize... egh, my bad, it was 800x450 just for the image purpose. XD

But I'd rather go with the 848x480 resolution... 1024x576 might be too big.

ntsc progaganda machine in full swing I see. 1024 X576 is the "proper" 16:9 resolution for PAL and since this will be for showing on a projector the larger resolution is better to have than the same quality as a smaller resolution.

I'm not very good in filtering, so I might just have said bullshit...

lies!
oh and .avs is dead again (amvapp3.0 hates me)I'll have it working again so I'll check back tomorrow i think (i'm getting faster at uninstalling all my editing software -_-)

@Nekmayumi- IVTC is not for PAL sources. I'll check out the pic you posted tonight as I can't see it at work the blocked the site x_x

[NekoMayumi]

@Nekmayumi- IVTC is not for PAL sources. I'll check out the pic you posted tonight as I can't see it at work the blocked the site x_x

Yeah I remembered that after I submitted the post. (would of edited but I can't seem to find an edit function on this forum >_>)

The pic isn't that great thanks to the scaling from I assume pre-scaled pic.

Also remove the . after my Crop function if you want to try that script. But like I said it isn't really any bit useful without an actual source clip to get an idea of what a good portion of the source is like.

It's just spatial, so I used the vaguedenoiser to have a temporal there.

Vaguedenoiser is a spatial denoiser not temporal.

Lastly here's a better version of my original try:
[/quote]

Script:
ImageReader("C:\PALisass.png").ConverttoYV12()
TBilateral(diameterL=9, diameterC=9, sDevL=4.4, sDevC=3.4, iDevL=12.0, iDevC=12.0, kernS=2, kernI=5)
FFT3DFilter(sigma=3.5, plane=3, bt=1)
DeHalo_Alpha(rx=1.0, ry=1.0)
Crop(4, 4, -4, -4)
src = last
EEDI2().TurnRight().EEDI2().TurnLeft().BicubicResize(width(src),height(src))

[mirkosp]

It's just spatial, so I used the vaguedenoiser to have a temporal there.

Vaguedenoiser is a spatial denoiser not temporal.

*checks the avtechbeta*
Omg, you're right! I misread the avtech so I thought it was temporal. Thanks to pointing that out, I won't be doing the same mistake anymore.