A Reconfigurable Architecture for 1-D and 2-D Discrete Wavelet Transform

Abstract

In this paper, we propose a novel architecture for DWT that can be reconfigured to be adapted to different kinds of filter banks and different sizes of inputs. High flexibility and generality are achieved by using the MAC loop based filter(MLBF). Classic methods, such as polyphase structure and fragment-based sample consumption, are used to enhance the parallelism of the system. The architecture can be reconfigured to 3 modes to deal with 1-D or 2-D DWT with different bandwidth and throughput requirements.