Symmetry Analysis of Face from a Video Image of 3D Point Cloud

Cleft lip surgery is performed several times. The problem is that the criteria for facial symmetry are unclear. We propose a method for the facial symmetry analysis using the 3D point cloud which is obtained as a video image. Pseudo video images are created by making multiple point cloud data with slightly different deformations. If we see the data in chronological order, the face appears to be gradually deforming. The symmetry is evaluated using the movement of the face landmarks. We detect face landmarks from an initial frame and perform matching technique on successive frames. After the matching process, we calculate a shift vector that describes the temporal change between the frames. Finally, we evaluate the symmetry by comparing its components on the left and right sides of the face. As an experimental result, the matching is performed with small error, and the symmetry is evaluated in a quantitative way.