Efficient 3D Crease Point Extraction from 2D Crease Pixels of Sinogram

In this study, we extract 3D crease points on the surface of an object from its X-ray CT projection images (sinogram) without reconstructing the surface mesh. The idea behind this is that most of the 3D creases are captured as crease pixels in the 2D projection images, so the creases may be reconstructed from the 2D crease pixels. So we first extract the crease pixels in each of the 2D projection images and apply reconstruction of the crease pixels to obtain 3D crease points. Connecting these crease points, a wireframe model is created. This method requires a small number of projection images, and thus makes it possible to reduce computational time. The error of the reconstructed wireframe model falls within half of the scaled pixel size.