Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

RGB based NNEDI3 #999

Open
Freestalkerdotfr opened this issue Sep 25, 2024 · 2 comments
Open

RGB based NNEDI3 #999

Freestalkerdotfr opened this issue Sep 25, 2024 · 2 comments
Labels
enhancement New feature or request

Comments

@Freestalkerdotfr
Copy link

Freestalkerdotfr commented Sep 25, 2024

Expected behavior 预期的功能

Hello, it would be nice to have an RGB based NNEDI3 upscaler. Is there a way that you do it or is there a way that I can do it myself ? I ask because the chroma part isn't properly upscaled, it result in color bleed, where in Libretro they get better chroma with the RGB based shader. https://docs.libretro.com/shader/nnedi3/ Thank you for your future inputs.
Edit : I can use the 256 neuron version
Edit 2 : I tried to use FSR and FSRCNNX but they have artifacts compared to NNEDI3 Luma, and NNEDI RGB will remove chroma artifacts where FSR based upscaler will make artifacts no matter what

Alternative behavior (optional) 近似的功能(可选)

No response

@Freestalkerdotfr Freestalkerdotfr added the enhancement New feature or request label Sep 25, 2024
@hauuau
Copy link
Contributor

hauuau commented Sep 26, 2024

I'm not quite sure if it's worth it. Weights in NNEDI3 are probably not optimal for RGB processing anyway.

But I managed to hack around a naive implementation. You could find it in the attached archive if you're willing to test it.

Note that processing performance takes at least 3x hit compared to the Luma version.
Shader compilation unfortunately also takes an almost literal eternity even compared to already slow compilation of the Luma version.

NNEDI3_RGB.zip

@Freestalkerdotfr
Copy link
Author

Thank you ! It took me a very long time to do the cache entry the first time I used the 256 neurons one, but the next time it will be instantaneous

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

2 participants