For signal and image analysis, the wavelet transform possesses the ability to extract details from the signal or image. Higher frequency components are extracted using finer time resolution, and lower frequency components are extracted using coarser time resolution. The transform decomposes a scanned signal into localized contributions for multiscale analysis. The authors present a general multiprocessor architecture for the efficient computation of signal decomposition with general wavelet bases. Their discrete wavelet transform (DWT) architecture is composed of a linear array of commercially available processors, which is easily reconfigurable for variable sized windows of data to be transformed. The use of commercially available processors replaces the costly special-purpose VLSI chip, and can be reprogrammed for other digital signal processing applications. Combining the filtering and decimation processes in the decomposition stage via a block state-space formulation achieves an efficient real-time implementation.
展开▼
