How can you get a 20Mpixel from a 5Mpixel sensor?
The 5Mpixels represent all the photosites, where 1/4th are red, 1/4th are blue, and 1/2th are green.
So when you convert all this data to a fully rgb pixels, you get -approx- 40bit fully colored RGB (5/4)Mpixels. (Jpeg reduce the color depth to 24bit along with "inflating" back to original resolution)
So in B/W you "give up" the color information, and just use the grayscale 0-1023 values for each pixel. Therefore you stay with 5Mpixels, unless you reduce the bit depth to extrapolate more resolution pixels.
Ain't that correct?
I'm not an expert so I may be wrong, but I think the idea is that in the sensor the physical pixels that are sensitive to each color may be located next to each other and not on top of each other. If so, there will be more resolution available for greyscale (but also geometrical problems if the layout isn't uniform, and the different colors obviously need proper weighting).
No... If the camera is 5Mpixels, then there are 5M photosites, thus 5M different light intensities can be recorded, no more!
In B/W picture you get EXACLY 5Mpixels with 10bit precise for each.
The way i see it, and i'll be happy if someone corrects me, it seems mathmatically you get a REDUCTION IN RESOLUTION in color mode, since 2x2 four different color pixels combine together to join their R, G, and B components to a fully colored pixel like we see on a screen.
That gives you quarter the resolution (in this case only 1.25Mpixels) but with 40bit depth for each of them (20bits for the green). JPEG then "inflates" back the data to 5Mpixel size in the cost of reducing the bit depth to 24bits, and still, when you zoom in to a 100% crop on an image, you can see that in it's fully true resolution, as you see it on the screen, it looks a bit soft, and i believe it is related to the 2x2 pixel joining issue!
