Thursday, June 23, 2011

What's Next ?

Here is a short story of a venture capitalist talking about The Making of Lytro, a company that wants to sell light field cameras. The main selling aspect is take a picture now and worry later. We know of other technologies that can even provide low cost modulation but it does not seem to have caught the imaging scene by storm.

In the imaging world, I know of two companies that are developing something based on compressive sensing .i.e. use a different types of modulation than the one (refractive) developed by Lytro to investigate niche markets:

Inview in particular is developing products based on technology developed at Rice where the modulation is performed with a DMD as in the single pixel camera. Works seems to be in the Infrared region where pixels are expensive. Applied Quantum Technologies seems to be focused on coded masks as developed in the CASSI system by the folks at Duke. That technology is showing interesting applications in hyperspectral imaging. Are there any larger markets that could be served with these compressive imaging technologies ?

That brings me to the next question: when I take a shot, do I generally worry about refocusing ? what do you worry about ?

On a totally different subject, Danny Bickson with his graphlab algorithm is making it to 4th place on the KDD Yahoo! contest. For those of you interested in using GraphLab (a way of performing parallel computation on many different cores), here is a small tutorial on Jacobi method in GraphLab. It would seem that one could think of using this technique for solving very large matrices when using SL0 for instance since the most expensive part of that algorithm is an SVD computation if I recall. Definitely something sparseman should consider if Google is interested in his Exacycle request

No comments: