Once again, a Salubrious Solstice to one and all, and my best wishes to you and hopes that you are enjoying this holiday season, for whatever definition of "holiday" works for you.
Last year, I described the wonders of "X" in a somewhat rarefied way. This year I'm going to be more down-to-Earth in my talk of the unknown. Namely, what sorts of digital camera improvements you may expect to find the [semi-mythical gift-giver of choice] putting in your [personal gift receptacle of choice] in the next double-handful of years.
This is not a wish list. Every one of these goodies is sufficiently mainline and part of the normal manufacturing track that I am certain they will all show up in your digital camera...unless current technology improves so much that it makes some of them unnecessary. That's the unknown part—either you'll get these goodies or you'll get something even better. It's horrible living with uncertainty, isn't it?
Here's the shape of shinies to come:
Back-illuminated thinned sensors: These have twice the collection efficiency of standard sensors, and they've been in common use in scientific instruments for many years. You haven't seen them in commercial cameras because thinning the substrate has been expensive. New manufacturing patents show how to do this on a production basis. It gets you a factor of two improvement in ISO, across-the-board, but it's especially important for small-pixel sensors.
Deep well insulating sensors: A problem with sensor arrays is electron leakage between the pixels. All sensors suffer from this. It reduces low-light sensitivity, decreases sharpness, and degrades color rendition. A new design adds a secondary well layer beneath the pixels that insulates them. Hard to estimate exactly how much this will improve things, but I would guess about half a stop in low light performance, plus better sharpness and color.
Non-Bayer filter arrays: Bayer filters are a good compromise when you have a moderate number of pixels. There are efficient at capturing spatial detail, and computationally simple. They have many disadvantages: poor overall light collection, horrid aliasing problems, and they shortchange blue sensitivity, especially bad for available-light work under normal indoor lights.
Other filter arrays solve all these problems. This is nothing new; Kodak and other companies have been doing this since the early days of high-quality digital cameras. As pixels approach the micron size (and there are very good reasons for them to do so, despite some of the naysaying you've read) quasi—random filter arrangements will improve sensitivity, sharpness, and reduce noise.
Adaptive pixels: Good digital cameras already capture an 11–12 stop luminance range, but there would be some real advantages to capturing a 20-stop range (total exposure freedom, for one thing). Adaptive pixels that alter their response depending on the intensity of light that hits them exist in laboratory sensors; they'll hit the mainstream within a decade, unless conventional designs get so much better that nobody cares.
GRIN optics: "GRadient INdex (of refraction)" lenses have an index of refraction that changes as you move out from the center of the lens. Scanners already use GRIN lens arrays to relay the image from the platen to the sensor. A few commercial camera lenses have already been designed with GRIN elements, but there are problems still to be solved in making large elements cheaply and characterizing them well in production. But, as with aspheric surfaces, which were similarly exotic decades ago, these will gradually penetrate optical designs.
Diffractive optics: Gratings embossed onto the surface of the lens can radically alter things like chromatic and spherical aberration, which are otherwise difficult to correct well. We already have limited diffractive optics. Like GRIN, this is going to improve a lot over the next decade and we're going to get very innovative designs.
Combination optics: What happens if you combine GRIN, diffractive gratings, and lens surfaces of arbitrary curvature? You get single lens elements that perform as well or better as your typical triplet. Imagine what you could do, in terms of lens weight, aperture, and size, if you could cut the number of lens elements by a factor of two or three and make each element thinner and lighter to boot.
This is all off-the-shelf in the next few-to-ten years. I didn't even get into the edgy stuff, like quasi-particle single photon spectral detection, metamaterials, and other exotica, whose future is entirely unknown.
Anyone who tells you that camera image quality is near to topping out simply has no idea of what's waiting in the wings. Salubrious Solstice!