We show that compressive sensing (CS) calculations are very ecient to reconstruct in 3D sparse objects whose 2D hologram has been recorded by digital holographic microscopy. The method is well adapted to image small scattering objects moving within a larger motionless object. This situation corresponds to red blood cells (RBCs) circulating in the vascular system of a zebrash (Danio rerio) larva. RBCs positions are imaged in 3D from a single hologram, while the RBCs trajectories, i.e. the perfused blood vessels, are imaged from a sequence of holograms. With respect to previous work (Donnarumma et al., Opt. express, 24, 26887, 2016), we get a gain of ∼ 500 in calculation speed.