MPEG-2 is a family of video and audio compression standards published as ISO/IEC 13818, whose video part (also issued as ITU-T H.262) became the codec behind DVD-Video, the first generation of digital television and much of professional standard-definition production. Among video codecs it is the direct predecessor of H.264 and the reference point against which later formats measure their compression efficiency.
The standard was finalized in the mid-1990s and combined block-based DCT coding with motion-compensated prediction in a form that hardware of that time could decode in real time. Its success came less from efficiency than from interoperability: one decoder chip could play discs, satellite feeds and cable channels alike.
How MPEG-2 compresses video
MPEG-2 removes spatial redundancy inside a picture and temporal redundancy between pictures. Each frame is split into 16×16 macroblocks, and each macroblock is further divided into 8×8 blocks that are transformed with the discrete cosine transform. The resulting coefficients are quantized, scanned in zigzag order and packed with variable-length codes, so smooth areas cost very few bits.
Temporal redundancy is handled by motion compensation with half-pixel precision. The encoder searches a reference picture for the best match of every macroblock, sends a motion vector and codes only the prediction residual. Unlike H.264 and later codecs, MPEG-2 has no in-loop deblocking filter, which explains the visible blocking when the bitrate is too low.
Pictures are arranged in a group of pictures (GOP) made of three picture types. The table below summarizes their roles.
| Picture type | Prediction source | Relative size | Role |
|---|---|---|---|
| I-picture | None, intra coded | Largest | Random access point and error recovery |
| P-picture | Previous I or P picture | Medium | Forward prediction chain |
| B-picture | Previous and next anchor | Smallest | Bidirectional prediction, not used as reference |
Typical distribution GOPs are 12 to 15 frames long, for example the pattern IBBPBBPBBPBB. DVD-Video limits a GOP to 18 frames in NTSC and 15 frames in PAL regions, which keeps seeking and channel changes responsive.
Profiles, levels and interlace support
MPEG-2 uses profiles to define which coding tools are allowed and levels to cap resolution, frame rate and bitrate. A decoder that claims a profile and level combination must handle every compliant stream within those limits, which is the basis of its interoperability. The most common combination is Main Profile at Main Level, written MP@ML.
| Combination | Maximum resolution | Maximum video bitrate | Typical use |
|---|---|---|---|
| MP@ML | 720×576 | 15 Mbit/s | DVD-Video and SD broadcast |
| MP@H-14 | 1440×1152 | 60 Mbit/s | HDV camcorders |
| MP@HL | 1920×1152 | 80 Mbit/s | HD broadcast in ATSC and early HD services |
| 422P@ML | 720×608 | 50 Mbit/s | Studio contribution and IMX recording |
| 422P@HL | 1920×1088 | 300 Mbit/s | HD studio formats such as XDCAM HD422 |
Main Profile uses 4:2:0 chroma subsampling, while the 4:2:2 Profile keeps twice the vertical color resolution for keying and grading. MPEG-2 was designed in the era of interlaced television, so it supports field pictures, field-based motion prediction and field DCT, which progressive-only codecs such as AV1 do not offer.
Systems layer and delivery formats
Part 1 of the standard defines how compressed video and audio are multiplexed. The program stream suits error-free storage such as optical discs, and the transport stream carries programs in 188-byte packets that tolerate transmission errors. The transport stream outlived the video codec itself and still carries H.264 and H.265 in broadcast and HLS delivery.
Real systems set bitrates well below the level ceilings. DVD video reaches about 9.8 Mbit/s at most, a standard-definition broadcast channel commonly runs at 3 to 6 Mbit/s, and an ATSC channel carries a 19.39 Mbit/s transport stream shared by all its programs.
MPEG-2 compared with later video codecs
Every later codec in the family improves on MPEG-2 with finer prediction and better entropy coding. As a rule of thumb, H.264 needs roughly half the bitrate for similar perceived quality, and H.265 roughly halves it again, although the exact gain depends on content, resolution and encoder quality.
| Feature | MPEG-2 | H.264 | H.265 |
|---|---|---|---|
| Block structure | 16×16 macroblocks, 8×8 DCT | Macroblocks with 4×4 and 8×8 transforms | Coding tree units up to 64×64 |
| Motion precision | Half pixel | Quarter pixel | Quarter pixel |
| Deblocking filter | None in loop | In loop | In loop plus SAO |
| Entropy coding | Fixed VLC tables | CAVLC or CABAC | CABAC only |
| Decoding complexity | Very low | Moderate | High |
The weakness of MPEG-2 is also its strength: decoding is extremely cheap and the patents have expired, the last of the major ones by early 2018. That makes it a safe choice where legacy hardware and long-term archive readability matter more than storage costs.
Where MPEG-2 is still used and what to watch for
MPEG-2 remains in service in DVD authoring, legacy terrestrial and cable networks, archive collections and professional formats such as XDCAM HD at 35 Mbit/s. Many playout servers still ingest it, and many broadcasters keep MPEG-2 masters from the 2000s.
| Symptom | Likely cause | Remedy |
|---|---|---|
| Blocking in dark scenes | Bitrate too low, no deblocking | Raise bitrate or use two-pass VBR |
| Combing on motion | Interlaced source shown progressively | Deinterlace correctly on playback or transcode |
| Banding in gradients | 8-bit depth and coarse quantization | Add grain or use lower quantizers |
| Rejected DVD build | GOP or bitrate exceeds the specification | Encode with a DVD-compliant preset |
When transcoding archive material, preserve the field order and frame rate, and avoid re-encoding MPEG-2 to MPEG-2 repeatedly, because each generation compounds quantization artifacts.
A non-technical view
MPEG-2 is an older way of shrinking video so that it fits on a disc or travels through a television channel. It does not store every frame as a complete picture. It stores a full picture every half second or so. In between, it keeps only the changes, such as a moving car or a turning head.
Think of it as a reliable family car from the 1990s. Newer models go further on the same fuel. Yet the old one still starts every morning, and every mechanic knows how to repair it. That is why DVDs and many TV systems still use it.
A concrete example
A small film festival wants to sell a DVD edition of a 95-minute documentary edited in high definition. The editor exports a 720×576 PAL master and encodes it to MPEG-2 Main Profile at Main Level with two-pass variable bitrate, averaging 5.5 Mbit/s with peaks near 8 Mbit/s and a 12-frame GOP.
The video stream takes about 3.9 GB, leaving room for an AC-3 audio track and menus on a single-layer 4.7 GB disc. Test discs play on several older players without stutter, and only one dark night scene shows faint blocking, which a slightly higher average bitrate in a second pass removes. The result is a disc that plays reliably on almost any DVD player still in use.