Simulation study of a refined approach for specular surfaces reconstruction in deflectometry

Regularization is a necessary step in deflectometry method to solve the inherent ambiguity by incorporating additional information. When the detected surface is large, complicated or convex, regularization by stereo methods could only provide normal vector information within a small overlapping region. In this case, a refined approach for the estimation of normal fields of the non-overlapping surface points resulting from the overlapping data is considerable. In this paper, a new method based on the approximation that normal vector of one point is perpendicular to the vector of connecting points at either side is proposed to calculate the non-overlapping normal information. Then the surface shape is obtained by integrating the normal fields with least-square 2D integration method. Simulation work with noisy data on a convex specular surface verifies its feasibility and accuracy.