A Study on Frame Interpolation Using Motion Vectors for Moving Pictures

摘要

In this work, we investigate on frame interpolation using motion vectors for moving pictures. Motion vectors obtained by block matching between two existing frames are used to interpolate non-existent frames in the original video sequence. In our approach, we adopt (i) sub-pixel motion estimation, (ii) MAP estimation, and (iii) overlapped block matching (OBM) to reconstruct frames with high precision (quality). Simulation results show that sub-pixel motion estimation improves the performance of frame interpolation, but the effectiveness of sub-pixel motion estimation is remarkably promoted by MAP estimation. In addition, more precise sub-pixel motion estimation helps to reconstruct higher quality frames by MAP estimation. Furthermore, OBM contributes to constantly improve the performance of frame interpolation.