Wavelets, filterbanks, and the Karhunen-Loève transform

Most orthogonal signal decompositions, including block transforms, wavelet transforms, wavelet packets, and perfect reconstruction filterbanks in general, can be represented by a paraunitary system matrix. Here, we consider the general problem of finding the optimal P ×P paraunitary transform that minimizes the approximation error when a signal is reconstructed from a reduced number of components Q