Pujitha Gunaratne, Nihal Kodikara, Yukio Sato
This paper presents an effective approach to analyz e asymmetries in facial expressions in three-dimension using range data. Five basic facial actions are captured on ten different subjects and their shape information is recorded with corresponding texture image. To analyze the range surfaces, which have been deformed characteristically during each facial action, pre -designed symmetric generic face mesh is adapted to measured range data using the least squares approximation method. Then mirror image triangular patches in left and right sides of the altered symmetric mesh are analyzed for each facial action to determine the degree of deformation. A simple variation comput ation method for the triangular patches in 3D is presented for the analysis of deformation. To illustrate the robustness of estimation, we compare the results of normal subjects with a patient with facial nerve paralysis disorders. High accuracy range data were obtained for the robustness of analysis using two high - speed rangefinder systems and their distinct configurations are discussed wit h potential impact on asymmetry analysis.