In most traditional head-surgery navigation systems, a binocular infrared camera is used to track retroreflective markers fixed on the human head, registering the 3D medical image data of the head to the human body. Generally, the registration markers are manually installed at specific positions. As the marker positions affect the accuracy of registration, they must be carefully installed, which extends the preparation time and may cause invasive damage to the patient. Meanwhile, when the navigation system uses a screen display, surgeons must repeatedly move their gaze between the surgical site and screen. The screen display is non-intuitive and imposes a cognitive burden on the surgeon. To resolve these problems, we propose markerless registration method using a structured light camera that registers 3D medical image data to the human body. The structured light camera obtains a point cloud of the patient's face. The precise head pose of the patient is obtained through a fast and accurate registration pipeline comprising coarse and fine registration algorithms. The registration method is combined with an optically transparent head-mounted display to build an augmented reality system. During multi-user experiments, the average accuracy of target registration was determined as (2.61 ± 0.55) mm, the average Dice similarity coefficient between the virtual and real models was 0.9548, and the average time of the registration algorithm was 0.47 s. Experimental results demonstrate that proposed method quickly and accurately automates the registration process. The proposed method can provide physicians with accurate target spatial information while meeting the accuracy and time requirements of surgery navigation. Moreover, proposed method can be further developed for applications in clinical settings.