the future of lenses, part 2

Lens technology is just beginning to progress beyond what Benjamin Franklin had to work with. Diffractive, aspherical, high-refractive-index, vibration-reduction lenses are starting to hit the market. More recent developments are electro-optical liquid or liquid crystal lenses. While these address peripheral problems in lens design like chromatic aberration and excessive bulk, the core problem is image quality: i.e., sharpness and contrast. The bottleneck for decreasing sensel pitch is increasing lens sharpness without sacrificing contrast.

I believe glass lenses are a dead end. They will always be heavy and bulky. They are not going to get ten times sharper, at least not anytime soon. Rather than having one massive light collector, I'd suggest developing per-sensel microlenses to have adaptive focus.

Let's look at what a microlens currently does. If I understand it right, it takes the light hitting the entire transistor containing a sensel and focuses it in to hit the fractional area that actually senses light. If it focuses the light enough to fit within the sensor, the sensor array will record all light hitting it. If the focused beam is too large, some of the image will be lost. (Note that in the currently prevalent Bayer array layout, much more than this is already lost, and must be interpolated.)

In current digital cameras, the light hitting the sensor array is already focused by the macro lens (macro in the sense of non-microscopic, not in the sense of close-focusing). The trick is to make the microlenses able to focus (via electro-optical liquid crystal or liquid) on a microprocessor-determined part of the scene. It's like a bug's compound eye: many images, each seeing a slightly different part of the whole picture. Digital images are already mosaics, so that aspect wouldn't affect image quality.

Advantages:

• weight and size: camera need only be big enough for the sensor, flash, battery, storage medium, and viewfinder
• simplified camera UI: no focus, no aperture; SLRs would become like P&Ss
• zero linear distortion (or extreme distortion, if that's what you want - this is already fixable in Photoshop, given sufficient resolution)
• greatly reduced chromatic aberration
• depth of field choice moved to post-processing
  • allows arbitrary selective focus (a view camera, lensbaby, and more)

This also opens up some new features for cameras. Consider:

• Wrap sensels around the outside of a cylinder (or even a sphere), and you get instant panorama. Of course, the shutter design would be interesting, and you'd need a timer or remote to not show up in the picture.
• Spy cam: point in one direction, shoot off to one side.

Challenges:

• The aiming of each microlens must be much more accurate than the aiming of current macro or micro lenses. Having to calibrate your array at regular intervals would suck.
• I'm sure there are many more I haven't thought of yet.

The point is that camera technology has lagged behind most other consumer products on the road to digitization. That's why we're seeing such rapid advances today. It's long past time we got rid of the bulky SLR form factor, with its expense, steep learning curve, time-consuming adjustments, and lens changing hassles. As an infamous Stanford lecturer once said,

"It should all just work."