Structured light is the process of projecting depth-encoding features onto a surface and using a camera to build a 3D model of the surface. As 3D scanners make their way into more consumer electronics, the ability to quickly acquire 3D models has become more important. While 3D scanning has traditionally been either a slow process to acquire a high definition model or an inaccurate process to quickly grab a large model, we propose a novel implementation that concerns itself with accelerating the acquisition of 3D point clouds by pruning the search space to only objects that have moved since the last frame. By alternating between projecting a one-shot depth encoding pattern and white light, we can use the generate a motion mask using the white light frame and make the assumption that points not in motion can keep their previously decoded position. New locations will only be searched and computed for points that reside within the motion mask. This work is showcased and profiled in a software implementation running on a CPU as well as a CUDA implementation running on a GPU. This work shows significant improvements upon traditional structured light implementations for scenes with a moderate amount of motion in the camera field of view for many different classifications of motion, though these improvements are subject to diminishing returns as parallelization increases.