Sparse Global Matching for Effective Video Frame Interpolation with Large Motion

Our method introduces a sparse global matching algorithm to effectively capture large motion in video frame interpolation, by integrating global-level information to compensate for the limitations of local-level flow estimation.