Low complexity preprocessing for 2D discrete multiwavelet transform

Multiwavelet transform can be implemented by tree-structured matrix filter banks which operates on vector coefficients instead of the scalar ones. Therefore, preprocessing is usually required to extract vector input from the input signal for better performance. A 2D approximation preprocessing scheme for discrete multiwavelet transform is proposed in this paper. The multiwavelet system considered is the one constructed by Geronimo, Hardin and Massopust [2] (GHM). Compared with 1D approximation-based method used in [5], the proposed preprocessing results in similar transform energy compression ability with additional advantage of lower computational complexity.