Fast wavelet transform for Toeplitz matrices and property analysis (Q2508040)

The authors propose new fast wavelet transform algorithms for Toeplitz matrices. These ones are achieved by a compactly supported wavelet that preserves the character of the Toeplitz matrix after the transform, which is quite useful in many applications involving Toeplitz matrices. The first part is an introduction concerning Toeplitz matrices, systems and fast transform algorithms proposed in recent years. The second part gives a brief review of the biorthogonal wavelets and theirs properties. The third part presents the general wavelet transform algorithms together with several theorems elaborating the characteristics of transformed matrices. The first correlative transforming algorithm for Toeplitz matrices is named complete BDWT (CDBWT). The second one is called partial DBWT (PDBWT) for Toeplitz matrices, and different from the CDBWT, this one only performs DBWT on the low frequency part of the matrix instead of the whole matrix. Then each transform is applied to a Toeplitz matrix (when the wavelet is orthogonal) or a matrix constructed by four sub-Toeplitz matrices (when the wavelet is biorthogonal). The fourth section gives numerical results which suggest that the new algorithms have better compression performance than the widely used trigonometric transforms for Toeplitz matrices. \noindent The new methods can preserve the Toeplitz characteristic of the original matrix and can be performed repetitiously to simplify many problems efficiently.