Author Topic: AviDemux/VapourSynth  (Read 9556 times)

mean

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 10967
Re: AviDemux/VapourSynth
« Reply #15 on: October 13, 2015, 07:21:16 PM »
Confirmed, U & V are inverted

YAFU

  • Newbie
  • *
  • Posts: 29
Re: AviDemux/VapourSynth
« Reply #16 on: October 13, 2015, 07:33:02 PM »
Ok  ;)
So this would be the script to reduce noise with KNLMeansCL using OpenCL if anyone is interested:
Code: [Select]
import vapoursynth as vs
core = vs.get_core()
##
##ffmpeg with ffms2
src = core.ffms2.Source(source='/PATH_TO_VIDEO_HERE/input.mp4')
##

YUV = core.fmtc.matrix(clip=src, mat="601", col_fam=vs.YUV)
YUV = core.fmtc.resample(clip=YUV, css="420")
YUV = core.fmtc.bitdepth (clip=YUV, bits=8)

clip = core.knlm.KNLMeansCL(YUV, d=3, a=2, s=4, h=4.2, device_type="GPU")

clip.set_output()

This is the documentation for parameters:
https://github.com/Khanattila/KNLMeansCL/blob/master/DOCUMENTATION.md

I clarify that I do not know if the script is well designed, but it seems to work.
« Last Edit: October 13, 2015, 07:36:30 PM by YAFU »