Skip to main content
European Commission logo print header

Sampling and Reconstruction driven by Sparsity Models with Applications in Sensor Networks and Neuroscience

Final Report Summary - RECOSAMP (Sampling and Reconstruction driven by Sparsity Models with Applications in Sensor Networks and Neuroscience)

Modelling signals as sparse in a proper domain has proved fruitful in many signal and image processing applications. More recently, the notion of sparsity has lead to new sampling theories that have demonstrated that the prior knowledge that signals can be sparsely described in a basis or in a parametric space can be used to sample and perfectly reconstruct such signals at a significantly reduced rate.
However, despite its great potential, the sparse sampling theory was still somewhat ad-hoc and often used in discrete settings and with relatively limited impact in applications.

The major achievement of this project has been to provide a unifying mathematical framework for the sampling and reconstruction of continuous or discrete sparse signals. The framework is based on the understanding of the interplay between approximation theory and the signal acquisition process. By developing this connection, we laid out a theory that allows sparse sampling ideas to be applicable with any real acquisition device and on a vast class of real-life signals.

The most immediate impact of the new theories and methods developed in the project has been in the domain of neuroscience where we produced state-of-the art methods for calcium-transient detection from two-photon calcium imaging.
Other important concrete achievements include a method to enhance the resolution of natural images that can be readily used with a normal smart phone and a theory to solve inverse source problems for physical fields using sensor networks.