Geek alert: dual photography
May. 11th, 2005 03:31 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
lederhosenNow this (warning: 63 Mb mp4) has to be the coolest thing I've seen all week; see also accompanying paper (19 Mb pdf).
In summary:
If you have a projector shining light onto an object, and a camera picking up that light, there is a certain relationship between what the projector projects and what the camera picks up. You can measure that relationship by running through a sequence of 'test patterns' on the projector, and this will let you predict how things will look to the camera just by knowing what the projector is screening.
That's not too earth-shattering. Where it gets interesting is that you can then reverse that relationship and get the answer to the question: "If I switched the projector and the camera around, what would the image look like?"
For a concrete demonstration of this, watch until the end of the video. I'm not particularly surprised by the theory involved here; what impresses me is how well it's been put into practice. If anybody had asked me, I'd have said "would work in theory, but you'll never get good enough data to do that in practice."
In summary:
If you have a projector shining light onto an object, and a camera picking up that light, there is a certain relationship between what the projector projects and what the camera picks up. You can measure that relationship by running through a sequence of 'test patterns' on the projector, and this will let you predict how things will look to the camera just by knowing what the projector is screening.
That's not too earth-shattering. Where it gets interesting is that you can then reverse that relationship and get the answer to the question: "If I switched the projector and the camera around, what would the image look like?"
For a concrete demonstration of this, watch until the end of the video. I'm not particularly surprised by the theory involved here; what impresses me is how well it's been put into practice. If anybody had asked me, I'd have said "would work in theory, but you'll never get good enough data to do that in practice."
