Lightweight Verification Schema for Image-Based Palmprint Biometric Systems

Table 3

Comparison of the proposed method to some state-of-the-art methods using the fusion of hand geometry and palmprint texture features.

Authors	Key points	Accuracy	Cite
Barra et al.	Authors extracted 57 geometric features from hand (lengths, areas, angles, and ratios) and used Euclidean distance for classification.	93%	[49]
Fang et al.	Authors used morphological operations (e.g., thinning) in order to create the line edge map and implement the Hausdorff distance for classification. Research was performed on the own database.	Up to 95%	[45]
Kozik and Choraś	Authors extracted 15 geometric features, calculated the variance value for each of the image blocks and used Haar wavelet. For matching, they tested two algorithms KLT and PCA.	Up to 94%	[50]
Kumar and Zhang	Research performed on own database. Authors used DCT and hand geometry in features extraction step and multiple classifiers (e.g., kNN, SVM, and decision tree).	77%–98%	[51]
Ungureanu et al.	Authors implement various features extractors (e.g., CompCode, OLOF, and RLOC) and various matching methods (e.g., SVM and kNN). Research was performed on 5 different mobile devices.	56%–81%	[47]
Giełczyk et al.	Lightweight verification schema based on fusion of the features presented in this work.	91%	—