Palmprint authentication based on discrete orthonormal S-Transform

• We propose a palmprint recognition method using 2D discrete orthonormal S-Transform.
• The performance of the method is evaluated by six different measurements.
• The PolyU, CASIA and IITD datasets are tested to evaluate the efficiency of method.
• Obtaining EER less than 1% for all databases illustrate the efficiency of method.

Personal authentication based on palmprint features has been attracting extensive attention currently. In many palmprint based recognition approaches, major issue is textural analysis of palm. In this paper, we propose a novel and efficient texture based approach to palmprint recognition based on a 2D discrete orthonormal S-Transform called as the 2D-DOST. 2D-DOST is a new powerful tool for texture analysis which has been introduced in recent years. It can efficiently characterize the frequency contribution of image texture in the various bandwidths. In this work, First, 2D-DOST is applied to the palmprint to characterize the frequency content of palmprint texture. Palmprint features are obtained by computing the local energy of 2D-DOST magnitudes in different bandwidths. Then, ICA will be employed to the extracted features to reduce the dimensionality and eliminate the redundant features. The experimental results offer EER equal to 0.93%, 0.97% and 0.12% for IITD, CASIA and PolyU databases, respectively. The results demonstrate the efficiency and validity of the proposed method.

Figure optionsDownload as PowerPoint slide