A framework for image-based 3D scene reconstruction based on photometric, polarimetric, and real-aperture methods
Dr. Christian Wöhler
DaimlerChrysler Research and Technology, Ulm, Germany


In this talk a variational framework for image-based three-dimensional surface reconstruction by combination of photometric, polarimetric, and real-aperture methods is presented. An error functional composed of several error terms related to the measured reflectance and polarisation properties is minimised by means of an iterative scheme in order to obtain a 3D reconstruction of the surface. The analysis of two or more pixel-synchronous images acquired under different illumination conditions allows to take into account an arbitrary non-uniform surface albedo. As a regularisation constraint, shadow information may additionally be incorporated into this framework. Since the described algorithm requires an initial surface that is reasonably close to the final solution in order to avoid local minima, it is initialised with the result of a depth from defocus analysis of the scene. By evaluating the algorithm on synthetic ground truth data it is shown that the combined approach strongly improves the accuracy of the surface reconstruction result, compared to techniques based on either reflectance or polarisation alone. In the application scenario of industrial quality inspection, 3D reconstruction results for a raw forged iron surface are reported along with a comparison to independently measured ground truth data. The described algorithmic framework is furthermore applied to the astrogeological problem of image-based 3D reconstruction of lunar surface regions.