Periodic dyadic wavelets and coding of fractal functions (Q1941845)

The first author has defined periodic dyadic wavelets on the positive semiaxis based on Walsh polynomials and Walsh-Dirichlet type kernels. In this paper, the authors generalize this construction with a modified Walsh-Dirichlet type kernel by introducing a parameter in the kernel. The corresponding wavelet decomposition and restoration are derived. Three coding methods are used to code the Riemann, Weierstrass, Schwarz, van der Waerden, Hankel, and Takagi fractal functions, where mean squared error and entropy of coding are utilized to evaluate their performances. Computational results show that the periodic dyadic wavelets with optimal values of the parameter are better than the Haar and Db4 wavelets in some cases. In short, the work as a generalization is theoretically meaningful, and the potential of the wavelets application is preliminarily shown.