3-D Medical Image Interpolation via Multi-Resolution Directional Correspondence

In this paper, we present a novel 3-D image interpolation method with high-quality feature preservation and low computational cost. The optimal direction of each voxel on the slices to be inserted is found by minimizing the smoothed directional correspondence using the Markov random field optimization approach. A multi-resolution scheme is employed to further reduce the memory consumption and computational costs as well as improve the interpolation accuracy. Extensive experiments are performed on medical image slices to evaluate the proposed approach, showing significant improvements in both accuracy and efficiency, as compared with the traditional interpolation techniques.