Wavelet shrinkage denoising using paraunitary shift-invariant filter banks

A new design method for shift-invariant (SI) wavelet filter bank (FB) is proposed and applied to wavelet shrinkage denoising. The SI wavelet transform obtained using the proposed method has better SI property than the conventional wavelet transform, which is highly desired in wavelet shrinkage denoising. Furthermore, orthogonality and dyadic structure are preserved, and computational complexity is reduced compared with those using the existing approaches. FB design examples, denoising results, and comparisons to the traditional orthogonal wavelet FB are presented.