We introduce a structured low rank algorithm for the calibration-free\ncompensation of field inhomogeneity artifacts in Echo Planar Imaging (EPI) MRI\ndata. We acquire the data using two EPI readouts that differ in echo-time (TE).\nUsing time segmentation, we reformulate the field inhomogeneity compensation\nproblem as the recovery of an image time series from highly undersampled\nFourier measurements. The temporal profile at each pixel is modeled as a single\nexponential, which is exploited to fill in the missing entries. We show that\nthe exponential behavior at each pixel, along with the spatial smoothness of\nthe exponential parameters, can be exploited to derive a 3D annihilation\nrelation in the Fourier domain. This relation translates to a low rank property\non a structured multi-fold Toeplitz matrix, whose entries correspond to the\nmeasured k-space samples. We introduce a fast two-step algorithm for the\ncompletion of the Toeplitz matrix from the available samples. In the first\nstep, we estimate the null space vectors of the Toeplitz matrix using only its\nfully sampled rows. The null space is then used to estimate the signal\nsubspace, which facilitates the efficient recovery of the time series of\nimages. We finally demonstrate the proposed approach on spherical MR phantom\ndata and human data and show that the artifacts are significantly reduced. The\nproposed approach could potentially be used to compensate for time varying\nfield map variations in dynamic applications such as functional MRI.\n