A face and palmprint recognition approach based on discriminant DCT feature extraction

Area of Science:

• Computer Vision
• Pattern Recognition
• Biometrics

Background:

• Discrete Cosine Transform (DCT) and linear discrimination are fundamental in image processing.
• Existing methods face challenges in optimizing feature extraction for recognition tasks.
• Effective feature selection and extraction are crucial for high-performance biometric systems.

Purpose of the Study:

• To develop an advanced face and palmprint recognition approach.
• To enhance feature extraction by selecting optimal DCT frequency bands.
• To improve classification accuracy and reduce feature dimensionality.

Main Methods:

• Utilizing a 2D separability judgment to identify discriminative DCT frequency bands.
• Applying an improved Fisherface method for linear discriminative feature extraction.
• Employing a nearest neighbor classifier for final classification.

Main Results:

• The proposed method demonstrated superior classification performance compared to state-of-the-art linear discrimination techniques.
• Significant improvements in recognition rates for both face and palmprint datasets were achieved.
• Effective reduction in the dimensionality of the feature space was observed.

Conclusions:

• The novel approach effectively integrates DCT and enhanced linear discrimination for robust face and palmprint recognition.
• The method offers a significant advancement in biometric recognition by improving accuracy and efficiency.
• This technique provides a promising direction for future research in image processing and pattern recognition.