Thanks for posting that - pretty interesting science/math. The thing is, this isn't very applicable to standard photography. The method is that you reconstruct what the typical sensor pixels must be based on the random patterns. I don't know the math, but the accuracy would go up the more samples you collected. To get to the accuracy of a typical snapshot (say, JPG accuracy, which is actually very good) you'd need lots of samples.
Just for the sake of argument, lets guess that for decent accuracy you need 1/2 the number of samples, as final pixels. That's 6 million samples. And you now need the mirror device to capture accurately that many times a second (and ignoring exposure time issues with the single pixel sensor). Then of course, there's the processing time to determine the final image pixels, which is actually the root of the exercise and a pretty significant challenge. It seems like it's probably simpler and cheaper just to create a regular sensor and be done with it.
The final applications of the technology look like they're still several years away at least (right now, small images take several minutes to take, with lots of equipment involved). But in the video, they mention security and medical application as most likely, implying that the hyperspectral (more than visible light) capture possibilities are more important in their view. Here's a quote from Kevin Kelly, one of the developers, in a different article (http://www.physorg.com/news79019816.html) (emphasis mine) :
For some wavelengths outside the visible spectrum, it's often too expensive to produce large arrays of detectors. One of the beauties of our system is that it only requires one detector. We think this same methodology could be a real advantage in terahertz imaging and other areas.
Besides not understanding what they were saying, I understood it completely.
It seems they found another wall after they tried to jump the hurdle of more pixels vs small file size. I don't think the processing will ever be fast enough for a professional camera.
Interesting topic, thanks for posting!
Discussion
2 repliesIf this works out, imagine the detail from a 12 mp camera.
the fisch
http://www.youtube.com/watch?v=xvIHHK_X3B8
Just for the sake of argument, lets guess that for decent accuracy you need 1/2 the number of samples, as final pixels. That's 6 million samples. And you now need the mirror device to capture accurately that many times a second (and ignoring exposure time issues with the single pixel sensor). Then of course, there's the processing time to determine the final image pixels, which is actually the root of the exercise and a pretty significant challenge. It seems like it's probably simpler and cheaper just to create a regular sensor and be done with it.
The final applications of the technology look like they're still several years away at least (right now, small images take several minutes to take, with lots of equipment involved). But in the video, they mention security and medical application as most likely, implying that the hyperspectral (more than visible light) capture possibilities are more important in their view. Here's a quote from Kevin Kelly, one of the developers, in a different article (http://www.physorg.com/news79019816.html) (emphasis mine) :
It seems they found another wall after they tried to jump the hurdle of more pixels vs small file size. I don't think the processing will ever be fast enough for a professional camera.
Interesting topic, thanks for posting!