Super Scale 2x Enhanced / Neat Video

Hey guys!

So I’m playing around with the new Super Scale 2x enhanced and have a bit of a specific question…

If I’ve already done my noise reduction and sharpening in Neat Video, and I turn both features down to 0 in the slider in the Super Scale setting, is the super scale still doing anything?

Hope I’m explaining this right

Also, am I correct in assuming Neat Video will be a more sophisticated method of applying NR and sharpening than through the super scale option

Thanks so much!

I have been using Superscale 2x Enhanced on my production deliverables for several months with settings 0.1 sharpening and 0.5 noise reduction in conjunction with my final node in the Colour page node tree. When I was doing my testing, I found the default setting of 0.5 sharpening produced unattractive results but 0.1 seems fine.

In the same Project Settings I also use Custom Gaussian rather than my normal default of Sharpen. I found this gave me much better results when my deliverables included externally rasterized text graphics provided to me by one of the bands I record. Without the Gaussian algorithm the text was hideous.

I know for some people NeatVideo is the best thing since sliced bread but it basically uses a noise profile to suppress noise which is subject to over-smoothing.

And by the way, Resolve 19 also has a noise profiling option.

The currently best noise suppression for video in my opinion is still 3D BMD

For super scaling, I would recommend a minimum or no noise suppression and I would stay away for at least a hundred miles of any sharpening.

Thanks Rick!

And thanks Cary! I’m very familiar with the pros and cons of Neat Video, especially how it can overly smooth footage, but after spending a few months with it I am very confident with the results I can produce with this program (turning quality to high, increasing frames to 5, turning on adapt to changing noise, making the NR in the red channel as low as you can etc)

I’m more curious about how the two programs interact

Also, what is 3D BMD?

I should mention that my production deliverables are shot in raw 4K and 2K and within Resolve there’s no other resizing of the footage so the super scaling is always dealing with a base of actual camera data, not computed pixel values from resizing. Superscaling presumably of course does compute the in-between pixels.

Also, what is 3D BMD?

https://en.wikipedia.org/wiki/Block-mat ... _filtering

Also, what is 3D BMD?

https://en.wikipedia.org/wiki/Block-mat ... _filtering

BM3D be done temporally? I never tend to like the results of spatial denoising algorithms. Even in Neatvideo, I'll only do spatial chroma denoising if I have spatial on at all. Spatial luma is always off and temporal denoising is doing 95% of the work.

I'm assuming block matching can be done across frames.

Also, what is 3D BMD?

https://en.wikipedia.org/wiki/Block-mat ... _filtering

BM3D be done temporally?

Yes.

https://github.com/WolframRhodium/VapourSynth-BM3DCUDA

Use the "v" version the selected radius determines the number of frames to consider.

If I’ve already done my noise reduction and sharpening in Neat Video, and I turn both features down to 0 in the slider in the Super Scale setting, is the super scale still doing anything?

I guess NR and Sharpening in the Super Scale menu are just very simplified versions of the color page NR and sharpen options? Which means they shouldn't affect the Super Scale effect itself. The manual doesn't say much.