The theory of compressed sensing (CS) predicts that random (or pseudo-random) linear measurements together with non-linear reconstruction can be used to sample and recover structured signals in a compressive manner. Lots of previous results demonstrated the efficiency of CS in recovering 2D images acquired using dedicated CS devices (single-pixel camera, accelerated MRI, etc…). In this paper, we investigate how this framework can be extended to perform an efficient joint reconstruction of a sequence of time-correlated 2D images, using 3D total variation regularization. We also evaluate the performances of this framework on test sequences issued from the bio-imaging field.
No takes yet. Share an insight, caveat, or question.
Montagner et al. (2012) studied this question.