Photometric Stereo with Area Light for Object-Level Surface Reconstruction

초록

A traditional photometric stereo method estimates surface normals with point lights placed at infinity, suffering from geometric errors when the light has a planar shape and is near the object. Though the light sampling approach improves the accuracy, it yields a trade-off between the number of light samples and the precision. In this paper, we propose a photometric stereo method to estimate the surface normals of an object using planar lights installed near the object, which is rather practical. Our multiple planar light model allows planar light to have different light intensities and near objects. We leverage the closed form of the shading equation of planar lights to build the linear equation that constrains surface normals by photometric observations. We compare our method with other physical-based approaches in the synthetic sets and demonstrate its performance in terms of accuracy and precision. We also test our method in in-the-wild scenarios.

키워드