用单幅图像重建三维凸面和两类鞍面的方法

利用Eikonal方程的粘性解给出了基于单幅图像的凸面三维重建算法；定义了2类鞍面——加性鞍面和乘性鞍面．利用凸面法和凹面法从鞍面单幅图像重建2个抛物柱面，然后根据鞍面的结构特点将2个抛物柱面重新合成鞍面．文中算法从临界点开始计算，不要求边界条件，具有较高的精确度．最后通过实验分析了该算法对噪声的敏感性．

3D Reconstruction of Convex Surface and Two Types of Saddle Surfaces Based on a Single Image

With the viscosity solution of Eikonal equation, a method is proposed on the 3D reconstruction of convex surface based on ~n image. Then, the additive saddle surface and multiplicative saddle surface are defined. By utilizing the methods of convexity reconstruction and concave reconstruction, two parabolic cylinders may be reconstructed based on one saddle image, and according to the characteristics of saddle surface, a complete saddle surface could be composed by the two parabolic cylinders. The algorithms possess the feature of high accuracy and no edge restriction is imposed since the computation starts from the critical point. In the paper, the sensitivity of algorithms to noise is also analyzed through experiments.