Fast and Accurate Image Registration for 3D CT Images

초록

This paper proposed a Fast and Accurate Image Registration (FAIR) method for 3D rigid CT (Computed Tomography) medical volumes. FAIR method involves two steps. First, the initial transformation parameters are calculated by Principal Component Analysis (PCA). Then, Nelder-Mead simplex method is employed to optimize the initial transformation parameters with the use of Normalized Cross-Correlation (NCC) similarity measure as objective function. The performance of the algorithm is examined on six CT phantom data sets. The experimental results show that the algorithm provides a fair image registration results which has higher accuracy than principal axis methods and lesser computational time than voxel similarity based methods.