Geometric-feature-based approach to human face reconstruction with high measurement speed

Area of Science:

• Computer Vision and Graphics
• 3D Reconstruction
• Optical Metrology

Background:

• Accurate three-dimensional (3D) face reconstruction is crucial for various applications, including biometrics and virtual reality.
• Existing methods often require multiple images, specialized hardware, or additional objects, limiting their practicality and speed.
• Phase unwrapping is a critical step in many optical 3D measurement techniques, but it can be challenging to perform accurately.

Purpose of the Study:

• To develop a novel, geometry-based method for 3D face reconstruction.
• To enable high-speed reconstruction of dynamic faces without extra equipment or images.
• To utilize the stable geometric features of the nose for robust phase unwrapping.

Main Methods:

• A geometry-based approach for 3D face reconstruction is proposed.
• The nose region is identified as a stable feature for phase unwrapping cues.
• Phase unwrapping is achieved by comparing the reconstructed nose shape with its known geometry, determining the correct phase offset.

Main Results:

• The proposed method successfully reconstructs dynamic faces with high measurement speed.
• Only three phase-shifted fringe patterns are required per frame, significantly reducing data acquisition time.
• The use of nose geometry for phase unwrapping proved effective and robust.

Conclusions:

• This novel geometric method offers an efficient and practical solution for 3D face reconstruction.
• The technique is suitable for capturing dynamic facial changes due to its high speed and minimal requirements.
• The nose-based phase unwrapping strategy enhances the accuracy and reliability of the reconstruction process.