A method for time-resolved imaging of N-dimensional magnetic resonance (=MR) with the following steps: Acquisition of MR signals from a sample volume by parallel imaging, wherein N-dimensional data matrices (M1, M2, . . . MNt) in k-space is acquired undersampled from each receiver coil, wherein the acquisition of the MR signals is performed according to an acquisition scheme that is periodic over time and describes the time sequence of the undersampled data matrices (M1, M2, . . . Mn) and reconstruction of missing data points (FP) of the acquisition scheme using a set of coil weighting factors (CW, ) and using N+1-dimensional reconstruction kernels (RK, RK′ RK″) is characterized in that reconstruction of the missing data points (FP) is performed using a single reconstruction geometry, wherein each reconstruction kernel comprises an (N+1)-dimensional target area (TB), wherein all non-acquired data points (TP) are reconstructed within the associated target area (TB) using each reconstruction kernel (RK, RK′ RK″), and wherein the target area (TB) exhibits at least the extent (nR×mR) in the ky-t plane of the acquisition scheme. This can shorten the computation time for reconstruction and reduce reconstruction artifacts and optimize the signal-to-noise ratio.