Laser scanning is widely applied to reverse engineering. Nonetheless, the involvement of scanning equipment and laser probe, as well as the measurement environment itself renders the process more vulnerable to noise which will cause failures in boundary detection, data segmentation and reconstruction of smooth curves and surfaces, thus putting forward the demand for noise reduction. Among the methods used in noise reduction of laser scanning data, the angular method is particularly suitable for scan line data. However, this method as well entails threshold that largely relies on the experience of engineers, which may introduce unwanted uncertainty. In this paper, an algorithm based on self-estimated angular threshold for reducing noise of laser scanning data is proposed. Firstly, the factors affecting angular threshold are analyzed and a mathematical model for SAT is established. Then the concrete algorithm that consists of judgment of impulse noise, modification of the coordinates and average filtering is described. Finally, a simulation test and an experimental case are employed to evaluate the performance of the algorithm on noise reduction of laser scanning data.