Novel Frame Rate Up-Conversion Method Through Weight Matching Criterion and Motion Vector Refinement

This paper presents a novel frame rate up-conversion (FRUC) framework using regulations matching criterion. Motion estimation is one of the key elements in FRUC, and the regularization matching criterion using the difference of Gaussians (DOG) is proposed to improve the motion estimation accuracy. The proposed FRUC framework has three steps. First, the initial motion vector field is calculated through the unidirectional motion estimation by two-pass neighbor recursive search. Second, motion vector refinement is used to get the more reliable motion vector through the bidirection estimation and motion vector postprocessing. Finally, the intermediate frame is reconstructed by linear motion compensation interpolation or overlapped block motion compensation according to the matching energy. The experimental results show that the proposed method can achieve comparable performance to other competitive algorithms with low complexity.