Selur's Little Message Board

When downscaling 4K stuff, using DGDecNV resizer gives me twice the fps when encoding with NVEnc on the GPU like every other resizer (it's even 50% faster than NVidia's built in resizers) , whether it is a cpu or gpu filter. And the results are somewhere between lanczos1 and Spline16. Sharp enough for me, especially when CAS sharpening with sharpness 0.6-0.75 is used as last filter.

Also DGDecNV resizer seems to reduce block artifacts in dark areas very good (also similar to spline16 and lanczos1), without softening the whole picture that much like gauss or cas-scaled does. Luma averageing is definitely done better in blocky dark areas than with spline36, spline64, lanczos4 or Blackman4. Huge visible block artifacts in dark areas suck as much as 8 bit banding in my eyes.

Too bad, that DGDecNV resizer is integrated in the source filter and always has to be the first filter in the Avisynth/VapourSynth filter chain. Or is there a way to use it after other filters, too?

No, the resizer in DGSource from DGDecNV is fast since it's part of the source filer and saves having to copy between VRAM and RAM.
In Avisynht DGDecNV also offers additional filters:

• DGDenoise()
  
• DGSharpen()
  
• DGTelecide()
  
• DGDecimate()
  
• DGBob()
  
• PVBob()
  

Quote:Currently, the utility filters DGDenoise(), DGBob(), etc., are available only when DGDecodeNV is used in native Avisynth mode.

see: 'Vapoursynth Notes.txt' next to DGDecNV under the 64bit subfolders.

=> Even if the filter would be available in a later step, it wouldn't be as fast as it is now.

Cu Selur