Reconstruct the Back of 3D Face Model Using 2D Gradient Based Interpolation

The recent development of 3D sensing technology enables a number of consumer facing 3D cameras, such as Kinect, TrueDepth camera on IPhoneX etc., emerge. These cameras are much cheaper than conventional and professional 3D scanning devices, and thus they can be acquired by consumers easily. However, consumer 3D scanning applications bring a new set of challenges. One of the challenges is that it is difficult for consumers to obtain the full head model by self-scanning. The proposed algorithm in this paper aims at reconstructing 3D human back head model based on gradient filling method. Due to the lack of related researches, to be more specific, repairing large holes without extra information in the 3D scale, the problem is migrated to 2D scale by projecting 3D model to a spherical space. Then the depth value at each position in back head is calculated via gradient interpolation. The algorithm is simple and effective and it can reconstruct a model within seconds.