I note the self explanatory slide 61, whereby information extraction is made directly on the compressed measurement thanks to low rank rank technique such as the ones mentioned yesterday. This is pretty much akin to manifold signal processing. I like it but I have yet to see it in the context of hardware calibration. The system used in this example is featured in another paper:
Coded Aperture Design for Compressive Spectral Imaging by Henry Arguello and Gonzalo Arce. The abstract reads:
Compressive sensing (CS) is an emerging field that exploits the underlying sparsity of a signal to perform sampling at rates below the Nyquist-criterion. This article presents a new code aperture design framework for compressive spectral imaging based on the Coded Aperture Snapshot Spectral Imaging (CASSI) system. Firstly, the methodology allows the CASSI system to use multiple snapshots which permits adjustable spectral and spatial resolution. Secondly, the measurement codeword matrices are generated using a pair of model equations, leading to code aperture patterns that permit the recovery of specific spectral bands of a given object. The developed methodology is tested using a real data cube and simulations are shown which illustrate that one can recover arbitrary spectral bands with high flexibility and performance.
Let us note that this system is devised to make up for the shortcoming of CASSI not unlike what Michael Gehm's Imager featured recently. Let us recall that a DMD/DLP chip and the board that drives it ( a Beagleboard) can be had for less than $500.
I'll add this design shortly in the compressive sensing hardware page.
No comments:
Post a Comment