Abstract
Recently, wavelet transform has been largely discussed in literatures, and proposed for next generation of multimedia compression standards. In this thesis, we propose an efficient scheduling algorithm and its related VLSI architecture. Based on the concept of recursive pyramid algorithm, we schedule the processing of different decomposition-levels in an interleaved manner. With the consideration of boundary problem, the architecture is designed with symmetric boundary extension. It yet possesses some advantages such as high hardware utilization, fast computation and regular data-flow, etc. In addition, it forms a framework of transform processor that can incorporate DWT with DCT with minimum hardware redundancy.We also evaluate the performance of JPEG2000 and hardware-software systems by building a simulation environment that off-loads the task of wavelet transform to a co-processor. The experimental results shows that finite-word-length effect is negligible in JPEG2000 coding system and speedup of the wavelet transforms can dramatically raise the coding speed.