View Full Version : aWarpSharp problem
Raven2x
18th July 2006, 06:23
Wierd distortion on left edge.
http://img.photobucket.com/albums/v202/Raven2x/bluestreak.jpg
It does this alot.
Removing aWarpSharp fixes this.
Script:
MPEG2Source("C:\Episode.d2v")
Import("LimitedSharpenFaster.avs")
Bifrost(Interlaced=True)
Crop(12,0,-12,0)
Interp = SeparateFields().SelectEven().EEDI2(field=1)
Deinted = TDeint(order=1,field=1,edeint=Interp)
TFM(d2v="C:\Gargoyles.d2v",mode=6,order=1,PP=7,slow=2,mChroma=true,Clip2=Deinted)
TDecimate(mode=1)
gradfun2db()
FFT3DFilter(sigma=3,plane=4,bw=32,bh=32,bt=3,ow=16,oh=16)
Lanczos4Resize(640,480)
aWarpSharp(depth=16,cm=1)
LimitedSharpenFaster()
I dont feel like cropping it off after aWarpSharp either.
Also can anyone recommend me a Dot Crawl remover besides dedot.
lamer_de
18th July 2006, 09:02
I don't know a solution to your warpsharp problem, other than to maybe try another version of it (the jp warpsharp dll for example, or a newer/older version of awarpsharp). I would've guessed it has something to do with mod 16, but you apply it after the resize to 640x480, so I don't know ;) Or maybe the awarpsharp screws stuff up because Limitedsharpen also uses/requires another version of warpsharp?
Also, using gradfun for encoding is not exactly reccomended, as the compression afterwards will very much void the results of the filter. It's basically meant to be applied during playback. You can read it up in the original release thread.
For removing dotcrawl, you can also try to upsize to 2x, apply sangnom or an other deinterlacer and resize back to the original size. Often the dots get recognised as "process-worthy" by those filters.
CU,
lamer_de
foxyshadis
18th July 2006, 09:05
awarpsharp has lots of bugs in its chroma sharpening. I think the picture has to be mod16 or mod32 to prevent most of them in mode 1, and mode 2 is just plain broken. Better to use the japanese warpsharp. (which, bizarrely, comes bundled with ivtc & unblending filters, and more.)
For dot crawl, try tcomb or a stronger fft3dfilter.
Raven2x
18th July 2006, 09:56
Found the japanese version, worked great.
Also, Tcomb worked great.
foxyshadis
18th July 2006, 10:16
http://www.geocities.co.jp/SiliconValley-PaloAlto/2382/warpsharp_2003_1103.cab
A zip version is at warpenterprises, iirc.
Raven2x
18th July 2006, 10:19
http://www.geocities.co.jp/SiliconValley-PaloAlto/2382/warpsharp_2003_1103.cab
A zip version is at warpenterprises, iirc.
Found it before you posted
:sly:
Works great too as far as I can tell
Chainmax
18th July 2006, 23:08
Raven2x: If DeDot isn't working for you, you might want to try using plane=3 in FFT3DFilter (so that it only removes rainbows, not noise) and using DeGrainMedian() for denoising, as IIRC Mug Funky once said that DeGrainMedian can remove some dotcrawl and even some blocking.
By the way, where are you loading BiFrost and the necessary DLLs for LimitedSharpenFaster?
awarpsharp has lots of bugs in its chroma sharpening. I think the picture has to be mod16 or mod32 to prevent most of them in mode 1, and mode 2 is just plain broken. Better to use the japanese warpsharp.
...
But can you exactly reproduce what aWarpSharp does with regular WarpSharp?
foxyshadis
19th July 2006, 01:54
I don't think so. I've never seen it make green edges! Unless you're talking about how it's supposed to look, in which case no, all four warpsharp algoithms (vdub's, marc's, the japanese, and ffdshow's) are subtly different, but they all share the same basic look and function. aWarp is just better optimized.
Chainmax
19th July 2006, 02:17
I've had aWarpSharp produce green splotches on the sides of the frame many times and never figured out why it happened. As for the rest of your reply, I always though that WarpSharp only did sharpening, whereas aWarpSharp only did line thinning.
foxyshadis
19th July 2006, 03:13
They all give it that thinned look, none of them are normal convolution sharpeners (which is why it's a good idea to use it in tandem with something like limitedsharpen). It's really easy to overdo it though.
The green splotch thing is the non-mod-16 (or 32) bug I mentioned earlier. At least, that's how I can always repeat it, other things might trigger it too. Marc was good at making stuff fast but bad at making them correct. =\
~bT~
3rd September 2006, 13:46
http://www.geocities.co.jp/SiliconValley-PaloAlto/2382/warpsharp_2003_1103.cab
A zip version is at warpenterprises, iirc.This is warpsharp.dll not aWarpSharp.dll or am I missing sumfin?
I need the japanese one too. Any help much appreciated.
unskinnyboy
3rd September 2006, 16:11
This is warpsharp.dll not aWarpSharp.dll or am I missing sumfin?
I need the japanese one too. Any help much appreciated.
Yes, it is warpsharp.dll, the Japanese version. This is the one to use.
~bT~
4th September 2006, 15:42
Yes, it is warpsharp.dll, the Japanese version. This is the one to use.
Thanks UsB.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.