Automatic face segmentation and facial landmark detection in range images.

Area of Science:

• Computer Vision
• Biometrics
• 3D Imaging

Background:

• Accurate face segmentation and landmark detection are crucial for robust 3D face recognition.
• Existing methods often rely on rich texture or multiple data types, limiting their application in depth-only scenarios.

Purpose of the Study:

• To develop an automated methodology for face segmentation and facial landmark detection using solely depth information.
• To integrate this methodology into a 3D face recognition system and evaluate its performance.

Main Methods:

• Face segmentation combines edge detection, region clustering, and shape analysis.
• Facial landmark detection utilizes surface curvature and depth relief curves to identify nose and eye landmarks.
• Experiments were conducted on the Face Recognition Grand Challenge (FRGC) and BU-3DFE databases.

Main Results:

• The proposed segmentation and landmark detection methods achieved high accuracy.
• Results surpassed those of current state-of-the-art approaches in the literature.
• The study demonstrated the positive influence of the segmentation process on the 3D face recognition system.

Conclusions:

• The developed methodology provides an effective depth-only approach for face segmentation and landmark detection.
• The system shows significant improvements in 3D face recognition, particularly when handling facial expressions.
• This work advances the field of 3D biometrics by leveraging depth information effectively.