A Massively Parallel EZW Processing Architecture for Real-Time Video Coding

摘要

Mobile multimedia communication has rapidly become a significant area of research and development constantly challenging the boundaries on a variety of technological fronts. In particular compressed video streams, a key component of typical multimedia applications, generally place a large processing burden on the targeted implementation hardware. The design of a single component, namely the Embedded Zerotree Wavelet coder, as found in a typical modem hybrid wavelet coder is presented in this paper. While the implementation of such a codec using a fast generic processor is possible, undesirable trade-offs in terms of power dissipation and speed must generally be made. The improvement in power consumption that is achievable through the use of a slow-clocked massively parallel processing environment, while maintaining real-time processing speeds, is exploited here. The design employs a pixel-wise self classifying technique to identify relevant EZW symbols and coefficients in a massively parallel manner. This paper therefore presents a massively parallel array based architecture which can perform the EZW.