Recent surge of interest towards increasing frame rate of existing videos to display the information with slow smooth motion has led to the demand for qualitative frame interpolation approaches that create spatially and temporally coherent intermediate frames. Traditional methods usually use only two adjacent frames to estimate the motion trajectories linearly, generally failing to overcome challenges like occlusion and non-linear motion. In this thesis, we introduce basic concepts regarding motion-compensated frame interpolation including optical flow, warping, splatting and inpainting and present an extension of linear forward and forward forward warping that uses arbitrary number of frames to approximate the motion trajectories in a polynomial way. We explore the difficulties of using optical flows over more distant frames and compare the use of sequentially warped optical flows with the direct variants. The best performing approach of this work named quadratic forward forward warping approximates forward and backward motion trajectories as parabolas, using three frames for every direction and requiring four frames in total, and utilizes a modification designed specifically to reduce warping artifacts. It achieves better results than the linear approaches on the high frame rate version of Sintel dataset proving that leveraging additional temporal information benefits frame interpolation. The use of more than three frames, however, is not beneficial and leads to decreasing performance.