3D Reconstruction by Perspective Shape from Shading Using Linearized Triangular Element Surface Model

Shape from shading (SFS) is an important domain in computer vision, also a difficult problem. An improved algorithm of shape from shading is proposed in this paper. Firstly, the presented algorithm is based on perspective projection rather than the common orthographic projection as in reality human beings are interpreting photographs perspectively. Secondly, the camera coordinate system is used, which is then transformed to viewer coordinate system. Thirdly, some constraints is enhanced to reduce the rotation distortion on the border of reconstructed object and the factor of brightness error to weaken the influence irradiance equation's nonlinearity on reconstructed errors is adopted. Eventually, the reconstructed surface is approximated by the union of triangular surface patches called triangular elements. The proposed algorithm is proved by experimental results of both synthetic images and real images, and the reconstructed images verify the performance improvement in terms of accuracy and speed