Learning the spherical harmonic features for 3-D face recognition.

Area of Science:

• Computer Vision
• Biometrics
• Pattern Recognition

Background:

• Traditional 3-D face recognition (FR) methods often rely on holistic approaches or local features from distinctive points.
• Capturing both gross shape and fine surface details of 3-D facial surfaces remains a challenge for existing FR techniques.

Purpose of the Study:

• To propose a competitive and effective method for 3-D face recognition (FR) using spherical harmonic features (SHF).
• To enhance FR performance by characterizing 3-D face models with energies from spherical harmonics at different frequencies.

Main Methods:

• 3-D face models are represented using a canonical representation: the spherical depth map.
• Spherical harmonic features (SHF) are calculated from the spherical depth map.
• Feature selection methods are employed to optimize SHF, considering their predictive contribution and improving efficiency.

Main Results:

• The proposed SHF method demonstrates effectiveness across three public 3-D face datasets (SHREC2007, FRGC v2.0, Bosphorus).
• The method shows robustness against variations in facial expression, pose, and occlusion.
• Feature selection improves predictive performance, leading to faster and more cost-effective 3-D face recognition.

Conclusions:

• The proposed spherical harmonic features (SHF) method offers a competitive solution for 3-D face recognition.
• This approach effectively captures detailed 3-D facial geometry, outperforming traditional methods in challenging conditions.
• The integration of feature selection enhances the practicality and efficiency of the 3-D FR system.