Numerical Condition of Discrete Wavelet Transforms

From MaRDI portal





In many applications biorthogonal wavelets have been used rather than orthogonal ones, since the latter might exclude other useful properties like symmetry in the case of compactly supported wavelets. Thus one would like to study stability of biorthogonal wavelets and obtain quantitative information about sensitivity to noise in the data or quantization. The author looks at discrete wavelet transforms (DWT) from the point of view of linear algebra. The recursive algorithm of a (fast) DWT, as well as its generalizations, can be described as repeated applications of block circulant matrices. Singular values of a block circulant matrix trigonometric series evaluated at certain points. The norm of a block-Toeplitz operator is then the essential supremum of the largest singular value curve of this series. For all reasonable wavelets, the condition number of a block-Toeplitz operator thus is the least upper bound for the condition of corresponding block circulant matrices of all possible sizes. The author develops and uses these results to study conditioning of biorthogonal wavelets based on \(B\)-splines.











This page was built for publication: Numerical Condition of Discrete Wavelet Transforms

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4377483)