Fig. 1. Distribution of edge points. (a) Distribution of edge points smaller than the mean value of curvature; (b) distribution of edge points larger than the mean value of curvature

Fig. 2. Schematic of the sparse grid in the lung

Fig. 3. Schematic of multi-scale registration framework

Fig. 4. Registration results. (a) Moving images in three anatomical directions respectively; (b) fixed images in three anatomical directions respectively; (c) registration result images obtained by using the proposed method in three anatomical directions respectively

Fig. 5. Comparison of difference graphs before and after registration using different methods. (a) Difference images between the fixed image and moving image before registration; (b) difference images between the fixed image and result image after registration using the smooth regularization term; (c) difference images between the fixed image and result image after registration using the TV regularization term; (d) difference images between the fixed image and result image after registration using the proposed method

Fig. 6. Comparison of execution time between uniform B-spline method and non-uniform B-spline method

Fig. 7. Comparison of TRE between uniform B-spline method and non-uniform B-spline method

Table 1. Comparison of TRE based on constraints of different regularization terms

Table 2. Comparison of TRE of different registration methods