Robust Registration of 3-D Ultrasound Images Based on Gabor Filter and Mean-Shift Method

Feng Cen, Yifeng Jiang, Zhijun Zhang, Hung-Tat Tsui, Tze Kin Lau, Hongning Xie

Abstract: A novel robust method is presented for the registration of 3-D ultrasound images. The proposed method improves the performance of the voxel property-based affine registration in two aspects. First, a set of wavelet-like Gabor filters is used to extract the texture and edge features of the voxels. By using these features, the smoothness of the similarity function in large scale can be improved. Furthermore, adopting edge information can improve the registration accuracy. Second, a robust maximization method based on the mean-shift algorithm and Powell’s direction set method is proposed. The implicitly embedded smoothing process of the mean-shift algorithm can effectively remove the local fluctuation of the similarity function and significantly improve the robustness of optimization. Experimental results demonstrate the robust and accurate performance of the proposed method in the registration of 3-D ultrasound fetal head images.