CORDIC-based architectures for the efficient implementation of discrete wavelet transforms

In this paper, efficient implementations of the discrete wavelet transform (DWT) are examined. A new architecture, based on single steps of the CORDIC algorithm with less operations compared to classical lattice or transversal filter structures is presented. In addition to the computational savings, a generalized implementation of a single CORDIC-based approximate rotation as a module generator is sufficient to automate the full custom layout generation of the arithmetic part of the DWT. An exemplary VLSI implementation for a typical image processing application is presented