Iterative photometric stereo with shadow and specular region detection for 3D reconstruction

Photometric stereo is a 3D reconstruction algorithm that uses the images of an object with different light conditions and its performance is affected by the shades and specular regions in the images. Especially, the use of Lambert reflectance model results in errors in the reconstructed surface normals. In this study, an iterative approach was used to generate masks corresponding to these problematic regions and the surface normals were reconstructed using a Lambert based algorithm that excludes these regions in calculations. A simulation model with varying albedo on a half sphere was constructed and the results for masked and non-masked cases were compared. The total error in calculated surface normals was reduced to 1.71% from 11.51% when the iterative photometric stereo algorithm was used.