Fast 3D reconstruction using one-shot spatial structured light

Structured light gains its popularity in 3D reconstruction applications due to its robustness against outliers. In the last few decades, a number of high-accuracy temporally encoded structured light emerged to solve the 3D reconstruction problem. However their applications are mainly limited to scanning stationary objects. When dealing with dynamic scenes and real-time data acquisition, one-shot spatial multiplexed structured light has the speed advantage. In this paper, we propose a fast 3D reconstruction method using one-shot special structured light. It works by projecting a static two-dimensional 8-color De Bruijn spatial grid pattern onto the scene, analyzing the deformation of the observed light pattern with respect to the projected one, and identifying their correspondence. Several local optimization strategies are used to offer a confident solution, including special vote majority for color detection and correction, and De Bruijn-based Hamming distance minimization to improve intersection neighborhood information. The effectiveness of the proposed method is verified through 3D reconstruction of a complicated bust.