Abstract: A registration algorithm for the spatial point sets in different coordinate frames was presented. Two steps, the coarse and fine registrations, were designed in the method. In the coarse registration step, the principle component analysis was used to estimate the three main axes for each point set. A space transformation was obtained by aligning the axes of the two point sets. Then the two point sets were roughly registered based on the space transformation. In the fine registration step, an improved iterative method for closest points was used to further optimize the registration locally, and to realize an accurate registration for two spatial point sets located in different initial directions. A phantom experiment was carried out to validate the proposed method. Results show that this method can align the main axes effectively in the coarse registration, and further optimize the results obtained from coarse registration in the fine registration. An accurate registration can be achieved between two point sets in different initial directions. The proposed method can improve the registration accuracy.