Comparison of Facial Region Normalization Techniques for Face Recognition

Key Points

• The aim is to compare various normalization techniques for facial regions in face recognition applications.
• Comparative analysis of four normalization techniques: wrapping, zero padding, context padding, and edge padding.
• Evaluation based on preserving the actual aspect ratio of facial regions during normalization.
• Testing the effectiveness of these methods with different algorithms.
• Zero padding yielded the best results without repeating facial edges or adding backgrounds.
• The impact of normalization techniques varies based on the specific algorithms used.
• The appropriate normalization method can significantly enhance system reliability.

Abstract