ISO/IEC 13818. A family of inter- and intra-frame compression systems designed to cover a wide range of requirements from ‘VHS quality’ all the way to HDTV through a series of compression algorithm ‘profiles’ and image resolution ‘levels’. With data rates from below 4 to 100 Mb/s, this family includes the compression system that currently delivers digital TV to homes and that puts SD video onto DVDs as well as putting HD onto 6.35mm videotape for HDV.
In all cases MPEG-2 coding starts with analyzing 8×8-pixel DCT blocks and applying quantizing to achieve intra-frame compression that is very similar to JPEG. This compression is referred to as I-frame only MPEG-2. Producing much higher compression involves analyzing the frame-to-frame movement of 16×16-pixel ‘macroblocks’ to produce vectors that show the distance and direction of macroblock movement. Their correctness is a factor of coders’ quality and efficiency. This vector data is carried in the P (predictive) and B (bi-directional predictive) frames that exist between I frames (see diagram). SDTV transmissions and DVDs typically contain two I-frames per second typically using about 4 Mb/s or less – a big difference from the 180 Mb/s of uncompressed SD video. The set of images between I-frames is a Group of Pictures (GOP) – usually about 12 for 576/50I and 15 for 480/60I transmissions. These are called ‘long GOP’. The GOP length can vary during transmission – an I-frame may be forced at the start of a new sequence, such as after a video cut, or other occasions were there is a big change at the input.
MPEG-2 12 frame GOP
*Note: for transmission the last ‘I’ frame is played out ahead of the last two ‘B’ frames to form the sequence I1, B1, B2, P1, B3, I1 B4, P2, B5, B6, P3, I2, B7, B8
Levels and profiles: MPEG-2 is a single compression standard that can operate on many different levels – picture source formats ranging from about VCR quality to full HDTV, and profiles – a collection of compression tools that make up a coding system. Current interest includes the Main Profile @ Main Level (MP@ML) covering current 525/60 and 625/50 broadcast television as well as DVD-video and Main Profile @ High Level (MP@HL) for HDTV. Besides the transmission/delivery applications which use 4:2:0 sampling, the 422 Profile (4:2:2 sampling) was designed for studio use and offers greater chrominance bandwidth which is useful for post production.
Blocking and ‘blockiness’: MPEG-2 artifacts generally show as momentary rectangular areas of picture with distinct boundaries. Their appearance generally depends on the amount of compression, the quality and nature of the original pictures as well as the quality of the coder. The visible blocks may be 8 x 8 DCT blocks or, most likely, ‘misplaced blocks’ – 16 x 16 pixel macroblocks, due to the failure of motion prediction/estimation in an MPEG coder or other motion vector system, e.g. a standards converter.
Audio: Digital audio compression uses auditory masking techniques. MPEG-1audio specifies mono or two-channel audio which may be Dolby Surround coded at bit rates between 32 kb/s to 384 kb/s. MPEG-2 audio specifies up to 7.1 channels (but 5.1 is more common), rates up to 1 Mb/s and supports variable bit-rate as well as constant bit-rate coding. MPEG-2 handles backward compatibility by encoding a two-channel MPEG-1 stream, then adds the 5.1/7.1 audio as an extension.