Tight wavelet frames via semi-definite programming (Q1958428)

The paper studies an optimization approach for multivariate tight wavelet frames in the shift-invariant setting, where the frame generators are defined in Fourier space. The results are based on the characterization of multivariate tight frames with maximal number of vanishing moments in the shift-invariant setting, called the oblique extension principle. Techniques of semi-definite programming allow to show that the existence of a tight frame is equivalent to the existence of a certain matrix from a cone of positive semi-definite matrices, whose entries satisfy linear constraints. A constructive method for computing the frame generators is thus given. The reduction of the number of frame generators in the univariate and the multivariate cases is also addressed. Examples for several subdivision schemes illustrate the theoretical results.