What Is Video Compression?

Video compression is the process of reducing the data volume of a video signal by removing or approximating information that is redundant or imperceptible to human viewers. Raw digital video requires enormous storage and transmission bandwidth: a single second of uncompressed 4K video at 30 frames per second and 8-bit depth occupies roughly 750 megabytes. Practical video systems depend on compression to bring this to manageable sizes, typically achieving ratios of 100:1 or more for web streaming while preserving acceptable visual quality.

Video compression is a subfield of data compression, applying techniques from information theory, signal processing, and perceptual psychology. It is closely related to image compression but must additionally exploit the temporal dimension of video, treating sequences of frames rather than individual images.

Spatial Compression

Spatial compression reduces data within a single frame by removing similarity between adjacent pixels. The central mechanism is the block-based discrete cosine transform (DCT), which converts an 8x8 or larger block of pixel values into a set of frequency coefficients. Because most natural images concentrate energy in low-frequency coefficients, the high-frequency coefficients can be coarsely quantized or discarded with little visible effect. Entropy coding methods, such as Huffman coding and context-adaptive binary arithmetic coding (CABAC), then compress the quantized coefficients further by assigning shorter bit patterns to more common values. Intraframe-coded frames, called I-frames, rely entirely on spatial compression and serve as random access points in a video stream.

Temporal Compression and Motion Estimation

Temporal compression targets the high degree of similarity between successive frames. A motion-compensated coder divides a frame into macroblocks, searches for the corresponding block in a reference frame, records the displacement as a motion vector, and transmits only the residual difference. This approach is described in detail in the ScienceDirect overview of video compression techniques, which documents how P-frames (predictive-coded) and B-frames (bidirectional-coded) achieve much higher compression ratios than I-frames alone. A group of pictures (GOP) structure interleaves I-, P-, and B-frames to balance compression efficiency against random access latency and error resilience.

Neural network-based video compression has emerged as an active research direction. A deep motion estimation study for parallel inter-frame prediction demonstrated that learned motion models can match or exceed the performance of hand-crafted block-matching algorithms on benchmark sequences, pointing toward end-to-end learned compression systems as a future alternative to transform-based codecs.

Compression Standards and Data Rates

The major video compression standards are products of the ITU-T VCEG and ISO/IEC MPEG committees working jointly or in parallel. MPEG-2 targets broadcast delivery at 4 to 15 Mbps. H.264/AVC achieves DVD-quality video at under 1 Mbps and remains the most deployed standard by device count. H.265/HEVC halves the bitrate of H.264 at equivalent quality and is the basis for 4K streaming. AV1, released in 2018 by the Alliance for Open Media as a royalty-free codec, achieves a further 30 to 50 percent reduction relative to H.265. A coding efficiency comparison across H.264, H.265, and AV1 published in IEEE conference proceedings quantified these gains under controlled encoder settings, providing a reference point used in codec selection decisions.

Perceptual quality metrics, including peak signal-to-noise ratio (PSNR) and the structural similarity index (SSIM), provide objective measures of compression quality, though neither fully predicts subjective viewer experience. Rate-distortion theory, from Shannon's information-theoretic foundations, defines the fundamental limit on how well any encoder can perform at a given bitrate.

Applications

Video compression has applications in a wide range of disciplines, including:

• Internet streaming and adaptive bitrate delivery
• Broadcast television and satellite systems
• Video conferencing and real-time communication
• Digital cinema and post-production workflows
• Surveillance systems with limited storage or bandwidth
• Mobile video capture and playback