"When an image is captured (effectively three 'photos', in this case, one for each of the RGB components), every part of the photo's viewing area is analysed, in turn, with the sharpest of the RGB images for each individual part determining the detail used (for that part), with the other images supplying appropriate coloration. Then it's onto the next part, in turn. The exact size of each 'part' is also a commercial secret but is likely to be in the order of 10 by 10 pixel squares. Here's an example of the output (taken on the Nokia E55):"

