Fast MR imaging techniques often exploit the redundancy present in an underlying MR image time series to compensate for k-space undersampling. When imaging motion using techniques like HARP, DENSE, and phase contrast (PC), it is the phase in static regions that is constant and therefore redundant. In this paper, we present a technique that first estimates the phase in the static portion of an undersampled MR image time series and then uses a partial Fourier reconstruction technique to combine phase in the static portion and under-sampled data to reconstruct the full image time series. The technique is illustrated using a computational phantom under-going simulated cardiac motion and imaged using the HARP protocol. Results demonstrate a gradual degradation of accuracy with loss of data due to undersampling, indicating that a 25% speedup in imaging time is possible for an image time series in which 50% of the pixels correspond to the object that do not move over time.