The noise does not contain a signal, does not dance over it, and is not detail. It is purely random fluctuations that are added to a signal.
If you have a few static frames and average them, you improve SNR by retaining the unchanged signal and having the purely random noise cancel itself out. Retaining noise itself is not useful.
I suspect the effect you might be seeing is either just an aesthetic preference for the original grain behavior, or that you are comparing low bandwidth content with heavy compression artifacts like smoothing/low pass filtering (not storing fine detail saves significant bandwidth) to high bandwidth versions that maintain full detail, entirely unrelated to the grain overlaid on top.
> If you have a few static frames and average them, you improve SNR by retaining the unchanged signal and having the purely random noise cancel itself out.
That's exactly the point of GP though. Even though each individual frame might be almost indistinguishable from random noise you can still extract patterns over time. This is also the case if you don't average the frames in software but let the viewer's brain do it. If you just remove all "noise" from each frame and then add random noise back those patterns will be lost.
In practice you won't have static frames but also movement so recovering the signal from the noise becomes a lot more complicated.
Anything with a pattern is by definition not noise, and the comment was that noise had signal. If you remove all noise, no signal or pattern is lost by definition.
However, the issue is that lossy compression removes various types of minute detail, smoothing surfaces to reduce the amount of data that has to be stored, be it noise "grain" or skin pores, according to compression settings. Storing the original noise as it was would basically make any compression impossible.
Denoising generally removes signal too. Removing noise and reconstituting similar noise to maintain the apparent qualities of an input can help compression, but you are also cutting out true details (typically fine detail).
The effect GP is pointing out is how denoisers damage detail, which is true. This detail can persist over multiple frames, which is why many denoisers include a temporal comparison component to mitigate the damage, but you still lose detail.
Definitely, but that's a result of compressing (with any algorithm) past a bitrate that would allow detail to be retained with high frequency details including noise being the first to go, not the result of noise having signal as was stated and more importantly unrelated to film grain synthesis which is about adding something new on the output side.
If you have a few static frames and average them, you improve SNR by retaining the unchanged signal and having the purely random noise cancel itself out. Retaining noise itself is not useful.
I suspect the effect you might be seeing is either just an aesthetic preference for the original grain behavior, or that you are comparing low bandwidth content with heavy compression artifacts like smoothing/low pass filtering (not storing fine detail saves significant bandwidth) to high bandwidth versions that maintain full detail, entirely unrelated to the grain overlaid on top.