A modification of the new edge-directed interpolation method is presented. The modification eliminates the prediction error accumulation problem with adopting a modified training window structure, and further extends the covariance matching into multiple directions for suppressing the covariance mis-match problem. Simulation results show that the proposed method achieves remarkable subjective performance in preserving the edge smoothness and sharpness among other methods in literature. It also demonstrates consistent objective performance among a variety of images.