A robust and high-performance shape registration technique using characteristic functions

We propose an innovative similarity registration method for volumetric shapes in this paper. This characteristic function-based method is intended to tackle the registration problem for the shapes containing sub-shapes in the presence of noise, and to strike a desirable balance between alignment performance and efficiency. In order to obtain the optimal parameters for scaling, rotation and translation in a reasonable time, radial moments and spherical coordinate system-based cross-correlation are exploited here. Moreover, an iterative method and principal component analysis are also employed to improve robustness of our algorithm. The shapes containing sub-shapes and the lung shapes from a CT dataset are employed in the experiments for validation. Compared with state-of-the-art algorithms, the characteristic function-based method manages to achieve excellent robustness at very low signal-to-noise ratio as well as superior registration speed, accuracy and stability in the medical shape data processing