On the construction of wavelets on a bounded interval (Q1906078)

The paper presents an approach to nonstationary multiresolution analysis and wavelet spaces on the interval \([-1,1]\). The method is based on the Chebyshev transform and corresponding shifts \[ (s_hf) (x) : = {1 \over 2} \biggl( f \bigl( xh - v(x) v(h) \bigr) + f \bigl( xh + v(x) v(h) \bigr) \biggr) \] with \(v(x) : = (1 - x^2)^{1/2}\). The space of level \(j\) of the multiresolution analysis is defined by \[ V_j : = \text{span} \{s_{h_{j,l}} \varphi_j : l = 0, \dots, N_j\} \] and the corresponding wavelet space by \[ W_j : = \text{span} \{s_{h_{j + 1,2l + 1}} \psi_j : l = 0, \dots, N_j - 1\} \] with \(N_j = d2^j\) and \(h_{j,l} : = \cos {l \pi \over N_j}\). As examples the authors consider for \(\varphi, \psi\) polynomials and transformed splines. Reconstruction and decomposition algorithms use fast cosine transforms.