Originally Posted by Fast Glass
From how I read the noise myth post, you will see increased noise, on a 100% crop, and not on a resize. I shall demonstrate my interpretation of that thread will little number buckets.
Lets say we have two sensors, the first 2x2 pixels, and the second 4 times the resolution, 4x4 pixels, in the same physical space. We've somehow convinced the photons to behave identically twice. The photons land such that each of the larger pixels, on the low res sensor, receive 8 units of light.
8 8
8 8
However, the light isn't really evenly distrubuted, and the smaller pixels, on the higher res sensor, show that. I've added some spacing to emphasis the where the original pixels were. Assume they're all nice and evenly spaced physically.
3 2 2 1
1 2 1 4
2 3 3 1
3 0 1 3
Each 2x2 block adds up to the same 8 photons that hit the same sensor area. While the total photon count is identical, to account for the lesser number of photons going to hit each pixel of the sensor, we need these values multiplied by 4 before being store in the JPEG.
12 08 08 04
04 08 04 16
08 12 12 04
12 00 04 12
<div>With such a small number of photons (low light, high ISO), the noise is quite apparent. A resize of the 4x4 image into a 2x2 image would smooth back to 8, effectively blending out the noise. The myth busting is essentially if you're against the 7D because 18MP will be noisier than the 10MP on your 40D, you're silly. You can resize to 10MP and it will look the same. In good light you won't see the noise, and you can take advantage of the extra resolution and crop, if you'd like.</div>
I could be wrong in how I'm interpreting the noise myth post. My interpretation still shows the noise being a mythical reason to avoid higher density sensors. It works out mathematically. It also shows that 100% there will be extra noise.