Getnative via AVS functions

[tormento]

I am starting to encode anime to their original resolution as I can save space and remove many bad resizing artifacts.

I use getnative (https://github.com/Infiziert90/getnative), but it's easily fooled when some filtering is used after the upscaling.

Plus, it needs VS to be used.

I went thru the FFT theory but, guys, I can't really read FFT graphs.

Is it possibile to have an AVS+ version, possibly smarter?

In short terms, I am looking a better way to find anime original resolution. I think it should apply to live action videos too.

[DTL]

You can test: downscale to some test size, upscale to input size, subtract original and upscaled to see how much difference.

[ChaosKing]

You can test: downscale to some test size, upscale to input size, subtract original and upscaled to see how much difference.

That's what getnative does :D

@GUID You need to test multiple frames. Sometimes credits have a different resolution. In this case you can use a mask to protect the text from upscaling (or downscaling).

[Kogarou]

I use getnative (https://github.com/Infiziert90/getnative), but it's easily fooled when some filtering is used after the upscaling.

Well, of course.

[tormento]

You can test: downscale to some test size, upscale to input size, subtract original and upscaled to see how much difference.
It means at least 540 trials... at least.

[DTL]

You can try iterative - first test 1/2 size. If it too small - test 1/2+1/4=3/4. And divide residual size to 2 to found direction to make next step.