Comparison Study of Extraction Accuracy of 3D Facial Anatomical Landmarks Based on Non-Rigid Registration of Face Template

• DOI: 10.3390/diagnostics13061086
• Publication Date: 2023-03-14T06:33:36Z
• Authors: Aonan Wen; Yujia Zhu; Ning Xiao; Zixiang Gao; Yun Zhang; Yong Wang; Shengjin Wang; Yijiao Zhao
• Keywords: Landmark

(1) Background: Three-dimensional (3D) facial anatomical landmarks are the premise and foundation of morphology analysis. At present, there is no ideal automatic determination method for 3D landmarks. This research aims to realize based on non-rigid registration algorithm developed by our team evaluate its landmark localization accuracy. (2) Methods: A scanner, Face Scan, was used collect data 20 adult males without significant deformities. Using radial basis function optimized algorithm, TH-OCR, (experimental group: TH group) MeshMonk (control MM group), a face template constructed in previous deformed registered each participant’s data. The realized according index 32 determined template. Considering these manually selected experts as gold standard, distance between automatically corresponding calculated “landmark error” effect feasibility (template method). (3) Results: mean error all groups 2.34 ± 1.76 mm 2.16 1.97 mm, respectively. middle better than that upper lower both groups. Further, center marginal part. (4) Conclusions: In this study, algorithms. There difference accuracy TH-OCR can meet needs oral clinical applications certain extent.