A method for under-sampling and compressed sensing of 3D spatially-resolved propagators is presented and demonstrated for flow in a packed bed and a heterogeneous carbonate rock. By sampling only 12.5% of q,k-space, the experimental acquisition time was reduced by almost an order of magnitude. In particular, for both systems studied, a 3D image was acquired at 1 mm isotropic spatial resolution such that 134,400 local propagators were obtained. Data were acquired in ~1 h and ~11 h for the packed bed and rock, respectively. It is shown that spatial resolution and under-sampling using this implementation retains the quantitative nature of the propagator measurement, and differences between implementation of this measurement in two and three dimensions are identified. The potential for 3D spatially-resolved propagators to provide new insights into transport processes in porous media by characterisation of the statistical moments of the propagators is discussed.