Compressive sensing in dynamic scenes

In recent years, the Compressive Sensing (CS) framework has received considerable attention. Most of its applications are found in static problems, such as the reconstruction of images from seemingly incomplete data. In this report we ask ourselves whether the CS framework can also be of use in a dynamic scene. In particular, we consider the task of tracking multiple targets. For this task, the class of Bayesian filters is optimal, but in some cases computationally expensive. We consider two alternative approaches, one that uses only CS, the other combines CS with a Particle Filter. These are hoped to improve on Bayesian filters in a situation where computational resources are constrained. For both approaches we propose alterations to the algorithms from the literature. We provide numerical results of the comparison between the proposed algorithms and the algorithms from literature they are based on. Furthermore we provide directions towards a comparison of the proposed algorithms and algorithms from the class of Bayesian filters.