About a week ago Nature.com posted this video about new research on imaging objects that are not in line-of-sight by using scattered light and an ultrafast camera. The researchers are from MIT, Harvard, and Rice University. So now if we can image hidden objects, what happens to the invisibility cloak?? :)
I watched the video yesterday and was excited to get into school and have access to the actual paper! It's fascinating what they've been able to do. Here are the parts of their "discussion" section which I find most interesting:
"Our reconstruction method assumes that light is only reﬂected once by a discrete surface on the hidden object without inter-reﬂections within the object and without subsurface scattering. We further assume that light travels in a straight line between reﬂections. Light that does not follow these assumptions will appear as time-delayed background in our heatmap and will complicate, but not necessarily prevent reconstruction."
The problem of reconstructing the hidden object using the scattered light is one example of what people in my field call an inverse problem. There are a lot of parameters involved-- the reflection surfaces, the textures of those surfaces, other cluttering objects. If you don't know the values of those parameters, then the problem becomes harder. Some of these problems have been looked at from a different perspective in the area of radar imaging.
One of the neatest things I have learned about recently is coded aperture photography. What that means is a type of filter is put on the aperture of your camera-- the filter is "coded" meaning that light from parts of the scene will be blacked out and parts will be allowed to pass (see this image for illustration). Coded aperture has been successful in improving imaging in many fields, including astronomical imaging and depth field imaging. In the Nature paper the authors also mention this as something that could be included, along with many other neat theoretical directions that are related to my work.
"A promising theoretical direction is in inference and inversion techniques that exploit scene priors, sparsity, rank, meaningful transforms and achieve bounded approximations. Adaptive sampling can decide the next-best laser direction based on a current estimate of the 3D shape. Further analysis will include coded sampling using compressive techniques and noise models for SNR and effective bandwidth. Our current demonstration assumes friendly reﬂectance and planarity of the diffuse wall."