An efficient architecture for orthogonal wavelet transforms

A novel architecture is advanced for the efficient high-speed computation of orthogonal one-dimensional two-channel discrete wavelet transforms. The structure is derived when the multirate operations (downsampling and upsampling) are staggered. Compared with the polyphase and lattice architectures the developed architecture requires the minimum number of multiplications. It preserves orthogonality even when the coefficients are quantized. For multilevel discrete wavelet transform, the proposed architecture allows balanced pipeline implementations.