Exact reconstruction algorithms for the discrete wavelet transform using spline wavelets (Q1907308)

Wavelet analysis has evolved in the last few years as a powerful method of describing signals at different resolutions. The wavelet transform is performed using translated and dilated versions of a single function which is called a wavelet. By making a discrete finite time signal periodic, it is shown that non-orthogonal \(B\)-spline wavelets can be used in a discrete wavelet transform with exact decomposition and reconstruction. A non-recursive algorithm using only finite impulse response filter with complexity \(O(N^2)\) is presented. The complexity is reduced to \(O(N \log_2 N)\) by using fast Fourier transforms. A fast algorithm is obtained by using recursive filters in the decomposition or reconstruction of the signal. The recursive algorithm has complexity \(O(N)\), and the same accuracy as the others. By allowing non-symmetric wavelets, an exact orthogonal reconstruction algorithm is shown, which has complexity \(O(N)\).