Earlier I wrote about desires for the next generation of DSLR camera and a number of readers wrote back that they wanted to be able to swap the sensor in their camera, most notably so they could put in a B&W sensor with no colour filter mask on it. This would give you better B&W photos and triple your light gathering ability, though for now only astronomers are keen enough on this to justify filterless cameras.

I'm not sure how easy it would be to make a sensor that could be swapped, due to a number of problems -- dust, connectivity and more. In fact I wonder if an idea I wrote about earlier -- lenses with integrated sensors might have a better chance of being the future.

Here's another step in that direction -- a "foveal" digital camera that has tiny sensors in the middle of the frame and larger ones out at the edges. Such sensors have been built for a variety of purposes in the past, but might they have application for serious photography?

For example, the 5d Mark II I use has 22 million 6.4 micron sensors. Being that large, they are low noise compared to the smaller sensors found in P&S cameras. But the full frame requires very large, very heavy, very expensive lenses. Getting top quality over the large image circle is difficult and you pay a lot for it.

Imagine that this camera has another array, perhaps of around 16 million pixels of 1.6 micron size in the center. This allows it to shoot a 16MP picture in the small crop zone or a 22MP picture on the full frame. (It also allows it to shoot a huge 252 megapixel image that is sharp in the center but interpolated around the edges.) The central region would have transistors that could combine all the wells of a particular colour in the 4x4 array that maps to one large pixel. This is common in the video modes on DSLR cameras, and helps produce pixels that are much lower noise than the tiny pixels are on their own, but not as good as the 16x larger big pixels, though the green pixels, which make up half the area, would probably do decently well.

As a result, this camera would not be as good in low light, and the central region would be no better in low light than today's quality P&S cameras. But that's actually getting pretty good, and the results at higher light levels are excellent.

The win is that you would be able to use a 100mm/f2 lens with the field of view of a 400mm lens for a 16MP picture. It would not be quite as good as a real 400mm f/2.8L Canon lens of course. But it could compare decently -- and that 400mm lens is immense, heavy and costs $10,000 -- far more than the camera body. On the other hand a decent 100mm f/2.8 lens aimed at the smaller image circle would cost a few hundred dollars at most, and do a very good job. A professional wildlife or sports photographer might still seek the $10K lens but a lot of photographers would be much happier to carry the small one, and not just for the saved cost. You would not get the very shallow depth of field of the 400mm f/2.8 -- it would be about double with a small sensor 100mm f/2 -- but many would consider that a plus in this situation, not a minus.

You could also use 3.2 or 2.1 micron sensors for better low-noise and less of a crop (or focal length multiplier as it is incorrectly called sometimes.)

One other benefit is that, if your lens can deliver it, and particularly when you have decent lighting, you would get superb resolution in the center of your full frame photos, as the smaller pixels are combined. You would get better colour accuracy, without as many bayer interpolation artifacts, as you would truly sense each colour in every pixel, and much better contrast in general. You would be making use of the fact that your lens is sharper in the center. Jpeg outputs would probably never do the 250 megapixel interpolated image, but the raw output could record all the pixels if it is not necessary to combine the wells to improve signal/noise.