Adaptive wavelet filter design for digital signal processing systems

Discrete wavelet transform has been used in many image/signal processing applications in recent years. However, the design of optimized and adaptive wavelet filter banks is still a significant research topic, specifically in image/signal compression. A number of wavelet-based advanced lossy compression algorithms provide high-fidelity reconstmction of input images at computationally intensive costs. The present work investigates the potential and the limitations of optimized adaptive design of two-channel perfect reconstmction filters when the signal in a channel is subjected to coarse quantization during the encoding process of such advanced compression algorithms.

A real-time optimal two-channel perfect reconstmction filter bank design algorithm has been developed and implemented in a digital signal processor. The algorithm has been used in a newly developed execution time reduction method to reduce the computational costs and data storage requirement of image compression algorithms. A reduction of execution time by two to three times has been achieved without adding appreciable distortion to the reconstmcted image.