Image registration using the shift-insensitive discrete wavelet transform

DWT (discrete wavelet transforms) has become an attractive tool in image registration, however, it is shift- sensitive, which is mainly caused by the aliasing term in downsampling process. Some modified schemes of DWT, such as DTCWT (dual-tree complex wavelet transform), always lead to high computational complexity in registration process because of redundancy. In this paper, a new scheme of discrete wavelet transform is proposed to approximately eliminate the aliasing term while the wavelet representation remains redundancy-free, by which the shift sensitivity of wavelet representations of original signal ca be reduced efficiently. The registration results of two medical images illustrate the superiority of the proposed wavelet decomposition scheme.