←back to thread

252 points CharlesW | 1 comments | | HN request time: 0.21s | source
Show context
crazygringo ◴[] No.44459098[source]
This fails to acknowledge that synthesized noise can lack the detail and information in the original noise.

When you watch a high-quality encode that includes the actual noise, there is a startling increase in resolution from seeing a still to seeing the video. The noise is effectively dancing over a signal, and at 24 fps the signal is still perfectly clear behind it.

Whereas if you lossily encode a still that discards the noise and then adds back artificial noise to match the original "aesthetically", the original detail is non-recoverable if this is done frame-by-frame. Watching at 24 fps produces a fundamentally blurrier viewing experience. And it's not subtle -- on old noisy movies the difference in detail can be 2x.

Now, if h.265 or AV1 is actually building its "noise-removed" frames by always taking into account several preceding and following frames while accounting for movement, it could in theory discover the signal of the full detail across time and encode that, and there wouldn't be any loss in detail. But I don't think it does? I'd love to know if I'm mistaken.

But basically, the point is: comparing noise removal and synthesis can't be done using still images. You have to see an actual video comparison side-by-side to determine if detail is being thrown away or preserved. Noise isn't just noise -- noise is detail too.

replies(7): >>44459330 #>>44459689 #>>44460601 #>>44461005 #>>44463130 #>>44465357 #>>44467163 #
1. BoingBoomTschak ◴[] No.44467163[source]
Don't want to sound too snarky, but aren't you just saying that good denoisers must be temporal in addition to spatial? Something like an improved V-BM3D (cf https://arxiv.org/abs/2001.01802) and even traditional V-BM3D works fine even on non-Gaussian noise.

Together with external noise generation (something like https://old.reddit.com/r/AV1/comments/r86nsb/custom_photonno...) that'd support external denoising (pass the reference and denoised video to get the grain spec), the whole FGS thing would be much more interesting.