Video Frame Interpolation

Video frame interpolation fills the gaps between observed frames, enabling smoother playback, slow motion, and reconstruction from sparse or imperfect video. Common methods estimate motion with optical flow, warp and blend input frames, and refine the result; newer approaches also use event-camera cues, state-space sequence models, and pretrained video diffusion models. Researchers evaluate image fidelity and perceptual quality on standard interpolation benchmarks, while extending methods to large motion, long sequences, spherical video, and RGB-D data. A central challenge is balancing realistic detail and temporal consistency with accurate motion, low latency, and reliable results when correspondences are ambiguous or frames are far apart.