Log in

View Full Version : Best de-combing/de-blocking script for NTSC anime


AOmundson
20th August 2022, 22:37
I just got a new Yu-Gi-Oh DVD set I'm trying to rip, and I'm unsure which script to use to best remove these jagged pixelations around the character outlines (see below screenshots). I've seen from other encodes that these artifacts don't always exist, so I'm mainly stuck on how to modify the script.

Top: My Rip - Bottom: Alternate Rip

https://i.imgur.com/Btk0rsq.png
https://i.imgur.com/JmhOLLn.png

https://i.imgur.com/vbrNbih.png
https://i.imgur.com/eoGJEV3.png

https://i.imgur.com/kFNjNke.png
https://i.imgur.com/8UV5DeV.png

https://i.imgur.com/PLJQpZN.png
https://i.imgur.com/34RO8It.png

This is my current script for IVTCing them, but it still leaves those blocky artifacts post-processing.

LoadPlugin("D:\DGDecNV\DGDecodeNV.dll")
LoadPlugin("D:\DGDecNV\TDeint-v1.8\x64\TDeint.dll")
LoadPlugin("D:\DGDecNV\TIVTC-v1.0.26\x64\TIVTC.dll")
dgsource("__vid__")
TFM()
TDecimate()

mastrboy
21st August 2022, 13:29
Looks like the usual dotcrawl/rainbow issue commonly seen on anime dvds, tcomb usually does a nice job with those:
dgsource("__vid__")
Tcomb(mode=2)
TFM()
TDecimate()

Though most decrawl/derainbow filters only work on static scenes there are some more "brute-force" ways to deal with dotcrawl in motion:
https://forum.doom9.org/showthread.php?p=1583181#post1583181
https://forum.doom9.org/showthread.php?p=1584186#post1584186

hello_hello
22nd August 2022, 00:14
I feel kinda silly posting this, as the script below really has nothing to do with dot crawl removal, and the definition of dot crawl means it's not a static thing, yet I'm only playing with a single frame, but I was working with a similar source recently and didn't have much luck with dot-crawl filters, so I started doing silly stuff. The source was HD so I wondered if downscaling before applying filters that aren't designed to fix dot crawl could fix it anyway, although this is SD which means there's not much downscaling wiggle room before picture detail is lost.
I wasn't doing anything else so I thought I'd give it a spin. It does need a de-rainbow filter added to the mix, but anyway, if nothing else works.....

ImageSource("D:\Btk0rsq.png").ConvertToYV24()
#ImageSource("D:\vbrNbih.png").ConvertToYV24()

Original = CropResize(704,528, 14,0,-4,0, InDAR=15.0/11.0)\
.ConvertToYV12()\
.subtitle("Original")

Output = BilinearResize(440,360)\
.QTGMC(InputType=1)\
.FineDehalo()\
.nnedi3_rpow2(2)\
.CropResize(704,528, 20,0,-6,0, InDAR=15.0/11.0)\
.FineDehalo()\
.FastLineDarkenMod4(Thinning=0)\
.MergeChroma(AwarpSharp2(Depth=8), AwarpSharp2(Depth=15))\
.CSMod(Strength=200)\
.ConvertToYV12()\
.GradFun3()\
.subtitle("Output")

#return Original
return Output

https://i.ibb.co/kq5mFKJ/A.png

https://i.ibb.co/0hBVqx3/B.png

https://i.ibb.co/J2GJt3T/AA.png

https://i.ibb.co/6RgqmKk/BB.png

AOmundson
3rd September 2022, 13:02
Looks like the usual dotcrawl/rainbow issue commonly seen on anime dvds, tcomb usually does a nice job with those:
dgsource("__vid__")
Tcomb(mode=2)
TFM()
TDecimate()

Though most decrawl/derainbow filters only work on static scenes there are some more "brute-force" ways to deal with dotcrawl in motion:
https://forum.doom9.org/showthread.php?p=1583181#post1583181
https://forum.doom9.org/showthread.php?p=1584186#post1584186

Just coming back to this thread to confirm that TComb did the trick. I also threw in DeDot for good measure, but I'm not sure if it actually helps any noticeable amount.