Learning Complete and Discriminative Direction Pattern for Robust Palmprint Recognition.

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Palmprint recognition commonly uses direction patterns, but existing methods rely on predefined filters, limiting accuracy and ignoring vital directional information.
• Noise in palmprint images and the challenge of extracting discriminative features hinder recognition performance.

Purpose of the Study:

• To propose a novel method for learning complete and discriminative direction patterns for enhanced palmprint recognition.
• To address limitations of existing direction-based methods, including reliance on prior knowledge and susceptibility to noise.

Main Methods:

• Extraction of complete local direction features (CLDF) and salient convolution difference features (SCDF) from palmprint images.
• Development of two learning models to derive sparse, discriminative directions from CLDF and uncover underlying structures in SCDFs.
• Concatenation of projected CLDF and SCDF to form a comprehensive and discriminative feature for recognition.

Main Results:

• The proposed method effectively learns complete and discriminative direction patterns.
• Experimental validation on seven palmprint databases and three noisy datasets demonstrates significant effectiveness.
• The approach shows improved recognition accuracy, particularly in the presence of noise.

Conclusions:

• The developed method offers a robust solution for palmprint recognition by learning superior direction patterns.
• This approach overcomes limitations of traditional methods, providing higher accuracy and better separability.
• The learned features are effective for palmprint recognition, even under challenging noisy conditions.