Even though startup finance absorbs a great deal (if not actually far too much) of my time, I continue in parallel to develop Nanodome’s security cameras – right now, for example, I’m developing the boot (startup) code to bring a new camera to life, so am currently bouncing back and forth between arm-linux-gcc and about ten USB development gadgets. Just a normal working day for most startup people. *sigh*
But I do think about other things too, and it struck me this morning just how curiously similar Pan / Tilt / Zoom (“PTZ”) cameras and human eyes are: for example, the narrow band of flexible cables connecting a PTZ’s camera to its wallside board is a lot like the human eye’s optic nerve.
Digital image compression has its equivalent in the eye too, insofar as the eye’s ~127 million photoreceptors (i.e. rods and cones) reduce down to only 1.25 million ganglions, which is already effectively a 100:1 compression ratio. The receptive field size of the ganglions also changes – smaller near the centre (the fovea), but progressively larger towards the periphery – giving a kind of spatial compression effect. All of this is so that the optic nerve can be as narrow as practical, yet still contain enough data to reconstruct the image in the brain.
Moreover, few people realise that the retina really doesn’t just capture images like a piece of film or an image sensor: it also processes them. Effectively, ‘on’ and ‘off’ ganglions convert what the retina picks up into a sparse edge-detected image which they then send down the optic nerve. The brain then processes this edge stream and silently reconstructs a gloriously detailed colour image back from it, which we then imagine that we saw (even though, technically speaking, we didn’t really): put like that, vision is an almost magical image pipeline built from an extraordinarily complex set of parts, which vision scientists are only just beginning to work out.
Of course, compared to the millions of years of mammalian retinal evolution, PTZs have only been around for 30-odd years: and so for all the marginal bells and whistles sales teams insist get added to differentiate their own company’s product lines, PTZ cameras still remain at heart brutally simple beasties (and, in many ways, not simple enough by half).
But there’s change in the air: the kind of ‘next generation’ camera my startup is building is dominated much more by software than by hardware. And because software evolution is normally that much faster than hardware evolution, the big industry trend looking forward is arguably going to be the rapid evolution of software-centric cameras as applied to specific problem domains. Errrm… which is exactly what my startup is doing, of course. 🙂
Given all this, I do wonder whether we would even recognize PTZs in 25 years’ time as PTZs, and (more to the point), whether we would even still recognize cameras from then as cameras. Hmmm… now there’s something to think about!