一种不受高光和阴影约束的形状恢复算法

针对高光和阴影同时存在时，用传统的4光源光度立体法进行三维形状恢复存在较大误差这一问题，提出一种4光源法向量选取算法．通过将不同像素恢复的法向量与镜面反射方向进行比较，并选取与之最接近的法向量为所求法向量进行形状恢复，避免了传统算法中用阈值剔除高光和阴影造成的误差，去除了高光和阴影对算法的约束，扩大了算法的适用范围，在高光和阴影未知的情况下，可以更加稳定地恢复出物体形状，而且大大提高了恢复的精度．仿真表明，该算法可以恢复不同颜色和材质的物体形状，与传统算法相比，在最坏的输入情况下，高度相对误差从40％～50％下降到5％左右．

Algorithm of Shape Recovery Without Highlight and Shadow Constraints

In order to reduce the error in three-dimensional shape recovery using traditional 4- source photometric stereo method when the highlight and shadow exist in the input images simultaneously, a new 4-source normal vector selection algorithm is presented. The shape recovery is carried out through comparing the recovered normal vectors from different inputs with the specular reflection direction and selects the one that is closest to the specular reflection direction as the required normal vector. It eliminates the errors caused by using the threshold to detect and exclude shadows and highlights in traditional methods. And it avoids the constraints of highlight and shadow on the algorithm, and extends the application scope. The improved algorithm can recover surface shape more accurately and stably without knowing shadows and highlights. The experiment results show that the proposed method can recover objects with different material and color. The relative height error is dropped from 40-50M to about 5% under the worst input condition.