Landmark-based homologous multi-point warping approach to 3D facial recognition using multiple datasets

Area of Science:

• Computer Vision
• Biometrics
• Medical Imaging

Background:

• Facial analysis research debates holistic vs. local feature acquisition.
• Advanced face recognition methods often utilize facial landmarks.
• Current 3D facial landmarking is mathematically complex and time-consuming.

Purpose of the Study:

• To propose a simplified and efficient method for 3D facial landmarking.
• To address the limitations of existing semi-landmark sliding tasks.
• To enable robust investigation of facial shape variations.

Main Methods:

• Developed a homologous multi-point warping technique for 3D facial landmarking.
• Utilized a template mesh as a reference object for artificial deformation.
• Applied semi-landmark sliding along tangents to minimize bending energy between template and target forms.
• Experimentally verified the method using 500 landmarks (16 fixed, 484 sliding) on three datasets.

Main Results:

• The proposed homologous multi-point warping method effectively performs 3D facial landmarking.
• The technique successfully handles artificial deformation and minimizes bending energy.
• Experimental validation demonstrated the method's applicability across diverse facial datasets (Stirling, FRGC, Bosphorus).

Conclusions:

• Homologous multi-point warping offers an efficient alternative for 3D facial landmarking.
• The method facilitates the investigation of shape variations in multiple facial datasets.
• This approach simplifies the complex workflow of current 3D facial landmark techniques.