[Method of constructing 3D facial smile simulation sequence data based on non-rigid registration]

To propose a novel method for constructing facial smile simulation sequence data based on static three-dimensional (3D) facial data captured at the start and end of smiling, and to preliminarily evaluate the accuracy and feasibility of the proposed method.

The 3D dynamic facial data of participants transitioning from a neutral expression to a maximum smile were captured using the 3dMD dynamic facial scanning system. A structured 3D face template was deformed and registered to both the smile starting and ending facial data using the Procrustes analysis non-rigid iterative closest point (PA-NICP) registration algorithm developed by our research group, obtaining two sets of structured homologous data. In MATLAB software, the vertex displacements between the corresponding points of the starting and ending homologous datasets were calculated, and intermediate transitional data with a consistent triangular mesh topology were generated through linear interpolation, thereby constructing the facial smile simulation sequence data. The real 3D dynamic facial data captured from the 3dMD system were used as reference data, and the simulation sequence data constructed in this study were used as test data. The 3D morphological deviations between the reference and test data at multiple time points during the smiling process were calculated to evaluate the accuracy of the constructed smile simulation sequence data.

The 3D facial smile simulation sequence data were successfully constructed for one male and one female participants. The average 3D morphological deviation for the simulated sequence of the male participant was (0.31±0.04) mm, and the average 3D morphological deviation for the simulated sequence of the female participant was (0.44±0.08) mm.

Based on the PA-NICP registration algorithm, the construction of facial smile simulation sequence data can be achieved. The intermediate transitional data can be parametrically generated and flexibly adjusted using interpolation functions, providing a novel method for 3D dynamic facial data generation that supports esthetic prosthodontic design, treatment outcome evaluation, and communication between clinicians and patients.