3D Reconstruction Of Gypsum Dental Casts

3D reconstruction is one of the important fields of computer graphics. In present, there are two main category of 3D reconstruction: one is dynamic reconstruction, using apparatus such as 3D laser scanner to gain the 3D coordinate of the points on object surface. The other one is static reconstruction, avail of object's 2D geometry and topological information to reconstruct its 3D information. Such as reconstruct from three views and reconstruct symmetry structure from perspective view. Comparing with dynamic reconstruction, static reconstruction require lower equipment and limitless space size in theory, but its arithmetic is more complex. 3D reconstruction from picture is an important 3D reconstructive method.This paper introduce 3D reconstruction of gypsum dental casts from sequence of images and describe images acquisition, camera calibration , three dimension reconstruction in detail.We use Zhengyou Zhang's approach to calibrate camera. This is a new and flexible approach and lies between the photogrammetric calibration and self-calibration. This approach supposes planar pattern which Z=0 in world coordinate, and analyses the camera parameters using linear model followed by a nonlinear optimization technique based on the maximum likelihood criterion. In the analytical process, lens distortion is considered. This calibration approach is flexibility, robustness and low cost. It advances 3D computer vision one step from laboratory environments to real world use.The images are taken from the object rotating in front of a stationary calibrated CCD TV camera. When object rotate an appointed angle, we take a photo until object rotated 360 deg. This process also can be regard as the camera takes many images of an object by rotating 360 deg.After got the sequence of images, we use two methods to reconstruct the dental casts.Volume model is one of them. We construct a volume space that contains the object and the volume space is decomposed into many small cuboids which are call volume. We project the cuboids to image. If the projective points in the background field, the cuboids are mark as white. If the projective points in object field, the cuboids are mark as black. After cuboids have projected in all image and marked, the black cuboids are the 3D entity of dental casts. Last, construct object's surface by Marching Cubes algorithm.The other one method is stereo vision. The each two adjacent images are regard as left and right image in stereo vision. After detected edge using Canny algorithm, we match feature point in the edge with constraint of epipolar line etc. Finally ,we use camera's parameter to calculate world coordinate of matched points and 3D reconstruct the edge of dental casts.