Open source library-based 3D face point cloud generation

Three dimensional (3D) face and body modeling is widely used in various fields such as plastic surgery, diagnosis of facial or body anomalies, 3D computer games and 3D simulation software. Since, commercial 3D face and body scanners are usually expensive, an alternative solution with lower cost is highly desirable. The objective of this study is to create 3D facial point cloud using Semi Global Image Matching method with minimum number of images utilizing a cost effective method. A non-metric Canon 600D camera with 18 megapixels resolution (3456 × 5184) and 60 mm macro lens have been used for face imaging that have been taken from a distance of 120 cm. Five faces have been modeled by the developed algorithm and scanned by David SLS-2 structured light system for accuracy assessment. Open source Cloud Compare software has been used for comparing the results of proposed method with the structured light system. The mean accuracy of five faces obtained as 90.5%. It has been observed that illumination conditions, uncontrolled movements of face or body, hair and eyebrow have negative impacts on the obtained results. The sufficiency of Semi global image matching method has been tested to create dense point cloud data from three stereo pairs for 3D facial modelling.