Question about diffraction limit

[Digigrapher]

You can't deny that a 20mp image has more (shot/photon) noise than a 10mp image! (because it simply resolves/reveals the noise/variation in the light itself better)

That actually proves the 20mp image has the exact same shot noise as a whole : if you look at the sum of all pixels, you get the total amount of converted photons.

Its statistics only depend on the total surface and the exposure. The number of distinct pixels does not matter in any way !

If you define the image's noise as the sum of all the pixels variances, you reach the same conclusion.

Some details : a single pixel's law is Poisson(K/n) with K depending only on sensor surface and exposure, and n being the pixel count.

Its noise (std deviation) is sqrt(K/n), so actually inferior to that of a single, wider pixel - but the pixels noise add in quadrature so the "total std dev" ends up being sqrt(K), regardless of n.

--
Samusan

Well, in that case I suppose that a 5mp image also has the same amount of detail as a 50mp image (we just can't see it!), since the lens projects exactly the same image on both sensors. No, what matters in photography is of course how well the noise and detail gets resolved, and a higher sampling frequency (more MPs) gives us more detail and more (visible!) noise in the image.

Not really. It all depends on what you mean 'more'. The per pixel noise has a higher amplitude, but extends over a smaller space, so it is not 'more'. In fact, it is 'the same' in terms of power. If you want to say that the noise amplitude is greater you would be right, but amplitude by itself is not the whole story.

Even given that the amplitude is higher at high sampling frequencies, a high sampling frequency allows the user to select lower frequencies, at which the amplitude will be the same as a native 'lower frequency' sensor. You get to choose which parts of your lunch you eat.