Originally posted by Ben_Edict No, I read the article as being quite supportive to my notion in the post you refer too.
Did you notice the title?
It says "More pixels offset noise!". Why would anyone title their article this way if they wanted to send the message that more pixels increase noise? A more elaborate version of that title would read "More pixels do not cause more image noise as the higher number of pixels offset the higher per-pixel noise".
Originally posted by Ben_Edict I'll quote a few key sentences from that article, though one should also be careful and mistrustful about the equations DXO give, because they obviously need to justify the processing they apply, as they don't make a living out of education, but out of the products they sell:
You should never simply believe anything, but always question what you read, agreed. Having said that, DxOMark's articles are based in solid Physics and I have yet to see one example where they were stretching the truth in order to sell their product. If anything, they should not educate people that more pixels do not equate to more noise, if they wanted to sell their denoising software.
Originally posted by Ben_Edict For a given exposure time, each smaller pixel receives four times less light than a large pixel—the equivalent of reducing exposure time by a factor of four.
That's correct but you are looking at a single pixel now. When comparing sensors with different pixel pitch (i.e., different number of pixels per unit area), it does not make sense to look at single pixels as you would be applying different magnification levels for examining each sensor.
You need to look at the full image, or at least at an area of the full image and allow both sensors to fill that area with as many pixels as they have available for that area. Then, and only then, you are comparing noise and other image properties at a fair basis. You are then applying the same magnification for your examination in both cases.
As the article explains, the four small pixels can be combined with binning to make one big pixel and that combined big pixel will have the same amount of noise as a real, physcial big pixel.
This is the real, key sentence in the article:
However, the four high-resolution neighboring pixels can be averaged out to form a low-resolution pixel.
Originally posted by Ben_Edict Also, it is hardly a measure to reduce noise by pixel binning.
Note that the effect of pixel binning is naturally achieved when printing an image. If you print the image from the high pixel count sensor to the same size as an image from the low pixel count sensor, inevitably many pixels of the high pixel count sensor will contribute to a space on paper that is influenced by only one pixel of the low count sensor. The same applies for scaling an image down for viewing on the screen.
Scaling down an image, whether for screen viewing or printing, implies noise reduction. You need to scale down high pixel count images more than low pixel count images. Hence the noise reduction necessary to counter the inevitable higher per-pixel-noise of high pixel count images is always taken care of in fair comparisons. And remember, looking at a single pixel is not a fair comparison as different magnification levels are used in this case.
Originally posted by Ben_Edict DxO introduce a kind of reference ISO and resolution (which is base for their own processing, I would guess).
The reason why they introduce the 8MP standard resolution is to allow fair comparisons between sensors with different pixel counts. The scaling required to normalise to an 8MP standard resolution takes care of the per-pixel-noise issue that high pixel count sensors otherwise would have.
Originally posted by Ben_Edict Relevant to our thread here would be to compare a FF camera to an APS-C modell of the same pixel count.
Once you accept that pixel count does not matter with respect to noise, you can rephrase this comparison to "compare an FF sensor to an APS-C sensor". The larger FF sensor only has an advantage, if you maintain the same light intensity per unit area. In this case, it will collect more total light which equates to a better signal to noise ratio.
If you are taking two shots, one with APS-C and one with FF, and use an equivalent f-ratio for your FF shot, but otherwise use the same shutter speed and the same ISO setting as for the APS-C shot, you will not get the same light intensity per unit area. The larger sensor only gives you the potential for a better signal to noise ratio and you'll have to invest by using a wider than equivalent f-ratio or slower shutter speed, or higher ISO setting. You can afford the latter with a good sensor easily as scaling down the FF image down to APS-C size, will take care of the additionally introduced noise.
P.S.: Pixel pitch is irrelevant for image noise in principle only. In practice there are limits to how small you can make the pixels without affecting the fill factor too much. But for our discussion and current pixel pitch values we can safely assume that smaller pixels do not have a disadvantage compared to bigger pixels as long as the overall sensor area is the same.