Vapoursynth-CCD

John Rambo · 12th December 2020, 18:54

https://github.com/DomBito/VapourSynth-CCD

Since I don't use Windows and Wine is a pain, I made this version of CCD for Vapoursynth using a bunch of Expr functions and shifted copies of the same footage.

It's not very fast, not even close to running the original one in Avisynth. But my friend tested on Windows and at least it's faster than loading Avisynth inside Vapoursynth and loading Avisynth's 'VirtualDub plugin loader' to run the original CCD inside VS.

The threshold might not scales the same as in the original and shownoise=True does not return a proper diff footage, but rather the chroma diff on a black clip (i.e. the luma channel is black in this one instead of gray).

If anyone more capable is willing to improving/properly redoing this, I would appreciate it.

feisty2 · 13th December 2020, 00:25

To write a convolution filter in C++ for VS, see: https://github.com/IFeelBloated/vsFi.../GaussBlur.hxx
I'm confident that it should not be any harder than python scripting

John Rambo · 15th December 2020, 03:21

Thank you, I will try it out as soon as i get home and post the results here.

Tima · 9th March 2021, 05:16

Any progress with the native implementation?

For me, current Python version is several times slower (and uses way more memory) than loading original VirtualDub plugin.

12th December 2020, 18:54	#1 \| Link
John Rambo Registered User Join Date: Jun 2020 Posts: 2	Vapoursynth-CCD https://github.com/DomBito/VapourSynth-CCD Since I don't use Windows and Wine is a pain, I made this version of CCD for Vapoursynth using a bunch of Expr functions and shifted copies of the same footage. It's not very fast, not even close to running the original one in Avisynth. But my friend tested on Windows and at least it's faster than loading Avisynth inside Vapoursynth and loading Avisynth's 'VirtualDub plugin loader' to run the original CCD inside VS. The threshold might not scales the same as in the original and shownoise=True does not return a proper diff footage, but rather the chroma diff on a black clip (i.e. the luma channel is black in this one instead of gray). If anyone more capable is willing to improving/properly redoing this, I would appreciate it.

13th December 2020, 00:25	#2 \| Link
feisty2 I'm Siri Join Date: Oct 2012 Location: void Posts: 2,633	To write a convolution filter in C++ for VS, see: https://github.com/IFeelBloated/vsFi.../GaussBlur.hxx I'm confident that it should not be any harder than python scripting

15th December 2020, 03:21	#3 \| Link
John Rambo Registered User Join Date: Jun 2020 Posts: 2	Thank you, I will try it out as soon as i get home and post the results here.

9th March 2021, 05:16	#4 \| Link
Tima Registered User Join Date: Aug 2004 Location: Russia, Novosibirsk Posts: 176	Any progress with the native implementation? For me, current Python version is several times slower (and uses way more memory) than loading original VirtualDub plugin.