Design and analysis of efficient reconfigurable wavelet filters

Abstract

Real-time image and multimedia processing applications such as video surveillance and telemedicine can have dynamic requirements of system latency, throughput, and power consumption. In this paper we discuss the design of reconfigurable wavelet filters for image processing applications that can meet such dynamic requirements. We generate several efficient hardware designs based on a derived family of bi-orthogonal 9/7 filters. An efficient folded and multiplier-free implementation of a 9/7 filter is obtained with the help of nine adders, which is a significant improvement over other existing approaches. We also propose an architecture that allows for on-the-fly switching between 9/7 and 5/3 filter structures. A performance comparison of these filters and their hardware requirements with other existing approaches demonstrates the suitability of our choice.