Adaptive Encoders: The New Generation of MPEG-2 Encoders

Abstract

Whereas current debate concerns more object-oriented technologies, such as the MPEG-4 standardization process, this paper adresses the state of the art of MPEG-2 technology. The MPEG-2 standard was adopted four years ago and is currently used for broadcasting digital television worldwide. Some comments and requirements from the first field trials require a new generation of MPEG-2 compression systems, and it was the objective of this paper to try to give some inputs for them. Whereas the first generations were mainly an optimization of the TM5 model with new motion estimators, new bit-rate control strategies and sometimes new mode selection algorithms, this new generation will be adaptive to the picture content and complexity. This adaptation concerns not only the input data (such as picture resolution, noise reduction) or the encoder itself (GOP structure, local adaptation, multipass encoding), but also the bandwidth (statistical multiplexing). From the results presented here, a 20 to 30% bit savings can be expected on average, and locally more than 50% (for special cases where the encoders fails, i.e,. noise, scene cuts, fading). After being in the cost and system software, the competition is now clearly on compression efficiency to add as many programs as possible into one single channel. The new techniques developed and introduced in this phase of optimization will certainly be exploited in future standards such as MPEG-4 or MPEG-7 where segmentation and scene characterization, respectively, will be key technologies.