Extending epipolar geometry for real-time structured light illumination II: lossless accuracy

Structured light illumination is a process of 3D scanning using a digital projector to project a series of striped patterns that sweep a target surface, and based on the warping of the stripes viewed by a camera, the shape of the target can be reconstructed. In the case of scanning stripes separately in both horizontal and vertical directions, algorithms have been proposed that achieve real-time reconstruction through look-up tables; however, these look-up tables implement the inversion of one traditional lens projection matrix. In this Letter, we propose look-up tables constructed using (1) both views of a camera and a projector and (2) basic arithmetic operations rather than complex matrix operations to significantly reduce the total number of computations used to reconstruct a point cloud. Experiments show that, with the same accuracy, the proposed tables improve the computation speed by a factor of 6.66×, from 31.00 to 206.61 fps.