Conclusions and Future work

      Compressive Sensing is a very interesting framework, which enables new imaging modalities, like the single-pixel camera. Maybe the random distribution of the cones at the back of the retina allows a compressive sensing of the world that can then be resolved by the brain at a much higher resolution than Nyquist would have predicted. The first objection to this hypothesis is that even if the cone distribution is random, it is fixed. It is also hard to believe that the brain is performing a l1 norm minimization. However, the compressive sensing paradigm does not depend on this specific reconstruction. An interesting simulation to perform would be to look at the respective reconstructions with a random measurement matrix and with a regular one, in a simulation that would include the optics of the eye (which ISET provides).

      As far as the rest of the visual system is concerned, we have seen that receptive fields might be what allows the brain to derive a sparse representation of natural images. It could also be the basis in which the sparse solution of the compressive sensing reconstruction problem would have to be looked for.

      At a higher level, compressive sensing is not limited to a measurement paradigm, and deterministic or dynamic extensions are promising research areas.