The shape-based motion estimation using frame differences for MCFI

We propose a new motion estimation algorithm for Motion Compensated Frame Interpolation (MCFI). In the proposed ME, a shape-based ME is performed to increase MV accuracy while decreasing computational cost. At first, we calculate frame differences using four frames to divide into four segmentations, moving objects, covered/uncovered area, and stationary background. Then, we generate two moving masks to use in the ME process. In addition, we propose interpolation method based on segmented information using the proposed moving mask. In experiments using benchmark test sequence, the proposed algorithm improves the average PSNR of interpolated frames by up to 2 dB compared to the conventional algorithm without complex post-processing algorithms.