In this paper, we propose a method of parallel processing of one-shot structured-light system to realize online 3D reconstruction of moving objects. The proposed method is based on a projector-camera system that reconstructs a shape from a single image where a static pattern is cast by a projector; such a method is ideal for acquisition of moving objects at a high frame rate. The structured-light pattern is a grid pattern that consists of sinusoidal wave lines of vertical and horizontal directions. Since the intersection points of wave lines are used as feature points and the lines give the connectivity of the intersections, the correspondence between camera image and projected pattern can be found by the energy minimization of the graph. The energy minimization is accomplished by belief propagation. We propose a method to improve the computational time for online 3D reconstruction by GPU-based parallel processing for the steps of the structured-light system. In the experiment, we show the proposed method realizes online 3D reconstruction.