Phase retrieval from fringe patterns is a primary procedure in fringe projection profilometry. Only accurate phase values result in three dimensions with certain accuracy. Phase shifting method plus temporal phase unwrapping approach provides not only the unwrapped absolute phase, but also the modulation map, background map, root mean square errors of least squares fitting, and phase relationship between two neighboring pixels, which can be used for the identification of phase invalidity. A practical phase retrieval frame work is presented to accurately calculate the absolute phase within reliable regions only, with which those unavailable phase points can be automatically identified with thresholds selection and criterion testing and then removed or interpolated according to applications. Experimental results show practical feasibility of the proposed framework.
Single-shot fringe projection profilometry based on deep learning and computer graphics
