Eaksarayut,W., Tunwattanapong,B., Sitthi-amorn,P., Chentanez,N.
We propose a technique to reconstruct a general 3D object using surface reflectance information from multiple
viewpoints. Our core optimization framework uses multi-view normal integration, which can recovers water-tight
surface of the object iteratively in a coarse to fine manner. The integration requires normal vector field from
multiple viewpoints, which we can derive from surface reflectance. We then handle the topological changes if
self-intersection occurs from the optimization. We also employ the idea of multi-resolution and weighted data
heuristic which helps dealing with noisy data and improves both accuracy and optimization time. Our experiment
shows that the framework is able to robustly recover 3D surface well with both synthetic and real data.