Subvoxel-accuracy surface detection method based on contour pre-segmentation for computed tomography images

Aiming at the precision and integrity problems of point cloud extraction in the detecting applications based on computed tomography(CT) images, this paper proposes a pre-segmentation based subvoxel-accuracy surface detection method with fine integrity and high precision. Firstly, the Otsu segment algorithm is adopted to obtain the initial sets of voxel-accuracy contour points for the CT image. With these sets as the coarse positioning contour, the complete region of interest(ROI) for subvoxel-accuracy surface detection is adaptively generated. Then, a surface voxel judging criterion is put forward based on non-maximum gradient suppression strategy, which avoids the gradient threshold selection dilemma. Finally, the positions of the subvoxel-accuracy surface points are determined based on 3D Facet model. Experiment results indicate that our method has a significant promotion in overcoming the contour loss and severe pseudo edges. The total positioning precision could be less than 0.2 voxels, and it could also obtain a computational speedup ratio above 3 compared with conventional methods.