Gabor feature based discriminative dictionary learning for period order detection in fringe projection profilometry

Fringe projection profilometry (FPP) is a popular method for accurate 3D model reconstruction. In a typical FPP process, a tedious phase unwrapping procedure is often needed to obtain the true phase information of the captured fringe images. However, conventional phase unwrapping algorithms often suffer from the ambiguity problems when the scene contains occlusions or sudden jumps in object's height profile. In this paper, we propose an efficient decoding strategy to overcome the ambiguity problems. A novel coded fringe pattern is employed and a Gabor feature based discriminative dictionary is used to estimate the unknown period order of the wrapped phase. Experimental results show that the proposed method can achieve an accurate 3D reconstruction of objects' model in various adverse conditions where traditional FPP methods often fail to perform.