Subspace design of compactly supported orthonormal wavelets (Q485214)

In the paper under review a method for designing compactly supported orthogonal wavelet systems is suggested. The method works in time domain. It is based on the singular value decomposition of the block matrices of the wavelet decomposition matrix. The coefficients of a low-pass wavelet filter are found as a solution of a linear system. The matrix of the system is parametrized by a right singular vector of the aforementioned singular value decomposition. Examples of such a parametrization for some standard orthonormal wavelets are given. This approach provides a closed form expression for wavelet filter coefficients. The suggested method is applied to solve two problems for wavelet design. The first one is how to obtain low-pass filter coefficients that satisfy additional linear constraints (for example, providing the maximum order of vanishing moments at zero point or obtaining a linear phase). The second one is a construction of matched wavelets. Examples illustrating the proposed framework are given.