Video Compression

Video compression studies how to represent moving images with fewer bits while preserving visual detail, smooth motion, or information needed for tasks such as recognition. Established hybrid codecs such as H.264, HEVC, and VVC combine prediction, transform coding, and entropy coding; learned codecs add neural motion estimation and learned probability models. Newer approaches use implicit representations, discrete latents, or diffusion and other generative priors to target perceptual quality at very low bitrates, while task-aware methods allocate bits for machine vision or adapt to network conditions. A central tension is balancing bitrate and perceptual quality against exact fidelity, temporal consistency, real-time decoding, and compatibility with standard hardware and codecs.