Inverting the Reflectance Map with Binary Search

Stereophotometric computer vision is based on the inversion of an equation system which represents the physical reflection phenomenon under different lighting conditions. In noisy conditions, the system has generally no solution, and the typical way to solve it is to find the solution which minimizes some error function. This paper proposes a new approach to deal with this system, using the so-called Geometrical Compatibility Function. The problem is turned into a root-finding in a one-dimensional space and the robustness is largely improved with respect to wrong reflectance model coefficients.