A method for generating a motion-compensated video frame based on four successive frames of an original video sequence is presented, where the four successive frames comprise frame A, frame B, frame C and frame D. The method comprises obtaining three successive motion estimation pairs, wherein a first pair comprises a forward motion vector field A_FWD from A to B and a backward motion vector field B_BCK from B to A, wherein a second pair comprises a forward motion vector field B_FWD from B to C and a backward motion vector field C_BCK from C to B and wherein a third pair comprises a forward motion vector field C_FWD from C to D and a backward motion vector field D_BCK from D to the C; and generating the motion-compensated video frame based on interpolating between the frames B and C, using the three successive motion estimation pairs.