Systems and methods of image reconstruction that can reduce the number of Fourier transform measurements required to obtain an exact or stable reconstruction of target images, using prior edge information obtained from a reference image. Full-sampled, 2 or 3-dimensional Fourier transform measurements are obtained for the reference image prior to performing a time series study. Based on the reference image sharing similar edge information with a time series of sparse-gradient images, the number of Fourier transform measurements required to reconstruct the target images can be reduced by using compressed sensing techniques to obtain the sparse-gradient images, and combining the edge information for the sparse-gradient images with the reduced number of Fourier transform measurements to obtain the exact or stable reconstruction of the target images, thereby permitting improved temporal resolution and/or extent of tissue coverage over the use of full sampling and conventional image reconstruction methods.
