Cross section binary coding for fusion of finger vein and finger dorsal texture

Abstract

In this paper, personal authentication by fusion of finger vein (FV) and finger dorsal texture (FDT) is presented. At first, a local enhancement method is proposed for FV and FDT images to enhance the vein vessel and finger dorsal texture line which can be called foreground lines (FLs). Then a novel cross section binary coding (CSBC) containing sixteen patterns is presented to find the foreground positions and quantify their cross-section profiles. CSBC utilizes the characteristic that for each point, the result of comparing four surrounding points with it reflects its location in the cross-section profile. Finally, a weighted fusion strategy is put forward. The experimental results declare its excellent performance.

Multimodal biometric recognition systems using deep learning based on the finger vein and finger knuckle print fusion

“…Our ResNet50‐Softmax architecture in the case of feature‐level fusion technique achieves better performances when compared to previous works. The proposed scores level fusion technique is also compared to Yang et al [25] proposed Weighted fusion and cross‐section binary coding methods. Weighted sum fusion achieves better results than those of previous works, thus the proposed methods are more performant.…”

“…In this point, researchers have proposed a huge number of multibiometric combinations. These combinations include for instance fingerprint and face [4], fingerprint and ECG [8], fingerprint and FV [22], palmprint and finger texture [15], FV and face [5], FKP and palmprint [23], FV and FKP [24, 25], FV, fingerprint and FKP [26], finger geometry, FKP and palm print [27] etc. In the case of fingerprint and face or ECG, the multimodal system needs several imaging devices inducing high complexity in the capturing process.…”

“…Although the obtained results are acceptable, the fusion strategy and fusion algorithms are complex and not reliable for real‐time recognition. Yang et al [25] designed also a multimodal biometric system based on score level fusion approach for the combination of FV and finger dorsal texture based on cross selection binary coding. In this work, they used the THV‐FVFDT2 database.…”

Feature Extraction for Finger-Vein-Based Identity Recognition

“…In 2016, Yang et al [ 71 ] used the Cross-Sectional Binary Code (CSBC) to extract the features from the finger vein and the finger dorsal texture and fuse them as one feature. In 2019, Liu et al [ 72 ] utilized Pixel Difference Vectors (PDVs) for feature extraction and then used the Anchor-based Manifold Binary Pattern (AMBP) for the feature learning process.…”