PVTree: Realistic and Controllable Palm Vein Generation for Recognition Tasks

TL;DR

PVTree introduces a two-stage palm vein generation framework that builds a 3D palm vascular tree via an improved Constrained Constructive Optimization and renders 2D vein patterns from multiple viewpoints. It enhances intra-class variation through a multi-view projection, depth-aware vein modulation, and an image-to-image translation step (PCE-Palm) to produce realistic vein images. Experimental results show PVTree-generated data can outperform real data for recognition under open-set conditions when identities are sufficiently large, and mixing synthetic with real data yields further gains. By bridging realistic vascular modeling with controllable diversity, PVTree reduces reliance on large real datasets and advances palm vein recognition training with synthetic data.

Abstract

Palm vein recognition is an emerging biometric technology that offers enhanced security and privacy. However, acquiring sufficient palm vein data for training deep learning-based recognition models is challenging due to the high costs of data collection and privacy protection constraints. This has led to a growing interest in generating pseudo-palm vein data using generative models. Existing methods, however, often produce unrealistic palm vein patterns or struggle with controlling identity and style attributes. To address these issues, we propose a novel palm vein generation framework named PVTree. First, the palm vein identity is defined by a complex and authentic 3D palm vascular tree, created using an improved Constrained Constructive Optimization (CCO) algorithm. Second, palm vein patterns of the same identity are generated by projecting the same 3D vascular tree into 2D images from different views and converting them into realistic images using a generative model. As a result, PVTree satisfies the need for both identity consistency and intra-class diversity. Extensive experiments conducted on several publicly available datasets demonstrate that our proposed palm vein generation method surpasses existing methods and achieves a higher TAR@FAR=1e-4 under the 1:1 Open-set protocol. To the best of our knowledge, this is the first time that the performance of a recognition model trained on synthetic palm vein data exceeds that of the recognition model trained on real data, which indicates that palm vein image generation research has a promising future.

Figures (7)

• Figure 1: The overall process of our method.
• Figure 2: Pipeline of the proposed method. Firstly, PVTree Model is used to generate a palm vascular tree, which starts from the trunk representing vein types that we have summarized, and an authentic 3D palm vascular tree is obtained through the Constrained Constructive Optimization(CCO) and Trajectory Simulation Mechanism(TSM). Then Multi-views project, crop, and blend it with the Bézier creaseszhao2022bezierpalm representing palmprint to obtain the palm vein patterns representing identity. Finally, by improving the noise sampling mechanism of PCE-Palm, diverse palm vein images are generated.
• Figure 3: Anatomical structure of palm blood vessels and their imaging under near-infrared light. Left: Vascular anatomical structure tan2020vascular; Right: A palm vein image from our dataset HFUT.
• Figure 4: Classification and construction of palm vascular tree trunks. Left: (a),(b),(c),(d) are the classification of vascular trunks based on observations of a large amount of real data; Right: Construction of type (a) trunk. The palm vein image comes from our dataset HFUT.
• Figure 5: The formation process of vascular tree’s branches.