10.1 Levels and profiles

MPEG-4, as with MPEG-2, has so many functionalities that users may only be interested in a subset of them. These are defined as profiles. For each profile, a number of resolution states such as bit rate and frame rate and pixel resolutions can be defined as levels. Since MPEG-4 is a very versatile video coding tool it has several levels and profiles, and every now and then many more are added to them. Profile in MPEG-4 is also synonymous with the support of a set of annexes in H.263. Some well known profiles with the associated commonly used levels are as follows:

Simple profile: this profile provides the simplest tool for low cost applications, such as video over mobile and Internet. It supports up to four rectangular objects in a scene within QCIF pictures. There are three levels in this profile to define bit rates from 64-384 kbit/s (64, 128 and 384 kbit/s for level-1, level-2 and level-3, respectively). The simple profile also supports most of the optionalities (annexes in H.263) that are mainly useful for error resilience transmission. In addition to I and P-VOPs (video object planes to be defined shortly) they include: AC/DC prediction, four motion vectors, unrestricted motion vectors, quarter-pixel spatial accuracy, slice synchronisation, data partitioning and reversible VLC. This profile can decode a bit stream generated by the core H.263.
Simple scalable profile: this adds the support for B-VOPs and temporal and spatial scalability to the simple profile. It provides services to the receivers requiring more than one level of quality of service, such as video over Internet.
Advanced real time simple: this adds error protection to the simple profile, through the introduction of the back channel. In response to a negative acknowledgement from the decoder, the encoder encodes the affected parts of the picture in intra mode. This profile improves the robustness of real time visual services over error prone channels such as videophone.
Advanced simple profile: this improves the compression efficiency of the simple profile, by supporting quarter-pixel resolution and global motion estimation in addition to B-VOPs.
Fine granular scalability profile: this is similar to SNR scalability, but the enhancement layers are represented in bit planes, to offer up to eight scalable layers. It is mainly used with the simple or advanced simple profile as the base layer.
Core profile: this adds scalability to still textures, B-VOPs, binary shape coding and temporal scalability of rectangular as well as binary shape objects to the simple profile. Its maximum bit rate for level-1 is 384 kbit/s and for level-2 is 1 Mbit/s. This profile is useful for high quality interactive services, as well as mobile broadcast services.
Core scalable visual profile: this adds object-based SNR, spatial and temporal scalability to the core profile.
Main profile: this supports for interlaced video, greyscale alpha maps and sprites. This profile is intended for broadcast services that can handle both progressive and interlaced video. It can handle up to 32 objects with a maximum bit rate of 38 Mbit/s.
Advanced coding efficiency: this profile is an extension of the main profile, but for bit rates of less than 1 Mbit/s. It adds quarter pixel and global motion estimation to the main profile to improve the encoding efficiency. However, it does not support sprites.
Simple studio profile: this only supports I-VOP pictures coded at very high quality up to 1200 Mbit/s. As the name implies it is designed for studio applications, and can support high resolution video for HDTV and digital cinema. Each I-VOP may have an arbitrary shape and have several alpha planes.
Core studio profile: this adds P-VOPs to the simple studio profile, to reduce the bit rate for very high quality video.