扩散模型用于指纹生成

… , starting with popular GAN architectures, and conducting fingerprint-to-fingerprint … , we focus on generating fingerprint patches and transforming fingerprints using diffusion models, a …

This paper reviews recent progress in fingerprint image generation using deep generative models, with a focus on Generative Adversarial Network (GAN)-based and diffusion-based approaches. Fingerprint data are essential for biometric recognition, but collecting large and diverse datasets is difficult, especially for latent fingerprints. Early methods based on physical or statistical modeling could not produce realistic textures or sufficient diversity. With the development of deep learning, GAN-based models such as FingerGAN, PrintsGAN, and lightweight GANs have significantly improved the realism of generated fingerprints by learning data distributions directly. These methods introduce techniques such as structural constraints, multi-stage generation, and improved loss functions to enhance image quality and stability. However, GAN-based models still suffer from problems such as training instability, mode collapse, and limited control over identity consistency. To address these issues, diffusion models have recently been introduced into fingerprint generation. By gradually denoising random noise, diffusion models can generate high-quality and diverse fingerprint images with better stability. Advanced diffusion frameworks further enable controllable generation, allowing users to adjust fingerprint attributes such as style, quality, and sensor type while preserving identity information. Overall, diffusion-based methods show strong potential to become the next generation of fingerprint synthesis techniques.

… In the aforementioned diffusion model, a pivotal aspect lies in training the noise prediction … the success of the Diffusion Probabilistic Model for latent fingerprint synthesis, with results …

High-quality, diversified, and large-scale datasets are crucial for creating reliable deep-learning models for biometric applications. Unfortunately, there is a shortage of well-labeled data. This paper introduces a text-conditional biometric imaging generation framework, addressing the complexities associated with multi-modality considerations. The proposed framework harnesses cutting-edge diffusion probabilistic models to produce multi-modal biometric images at high resolutions, seamlessly aligning with biometric language prompts. The experimental results unequivocally validate the efficacy of the proposed framework in generating a diverse array of highly realistic synthetic biometric images while consistently maintaining a commendable level of fidelity when juxtaposed with their respective reference datasets. The contributions of this study offer substantial potential for propelling advancements in biometric imaging research.

The advancement of fingerprint research within public academic circles has been trailing behind facial recognition, primarily due to the scarcity of extensive publicly available datasets, despite fingerprints being widely used across various domains. Recent progress has seen the application of deep learning techniques to synthesize fingerprints, predominantly focusing on large-area fingerprints within existing datasets. However, with the emergence of AIoT and edge devices, the importance of tiny partial fingerprints has been underscored for their faster and more cost-effective properties. Yet, there remains a lack of publicly accessible datasets for such fingerprints. To address this issue, we introduce publicly available datasets tailored for tiny partial fingerprints. Using advanced generative deep learning, we pioneer diffusion methods for fingerprint synthesis. By combining random sampling with inpainting diffusion guided by feature keypoints masks, we enhance data augmentation while preserving key features, achieving up to 99.1% recognition matching rate. To demonstrate the usefulness of our fingerprint images generated using our approach, we conducted experiments involving model training for various tasks, including denoising, deblurring, and deep forgery detection. The results showed that models trained with our generated datasets outperformed those trained without our datasets or with other synthetic datasets. This indicates that our approach not only produces diverse fingerprints but also improves the model’s generalization capabilities. Furthermore, our approach ensures confidentiality without compromise by partially transforming randomly sampled synthetic fingerprints, which reduces the likelihood of real fingerprints being leaked. The total number of generated fingerprints published in this article amounts to 818,077. Moving forward, we are ongoing updates and releases to contribute to the advancement of the tiny partial fingerprint field. The code and our generated tiny partial fingerprint dataset can be accessed at https://github.com/Hsu0623/Inpainting-Diffusion-Synthetic-and-Data-Augment-with-Feature-Keypoints-for-Tiny-Partial-Fingerprints.git

Fingerprints have been crucial evidence for law enforcement agencies for a long time. Though the rapidly developing deep learning has dramatically improved the performance of the latent fingerprint recognition algorithm, a fully automated latent fingerprint identification system is still far from meeting actual needs. One major issue is the lack of publicly available latent fingerprint databases. Recently, diffusion probabilistic models have emerged as state-of-the-art deep generative methods for image synthesis. These models have better distribution coverage and less mode collapse than the popular Generative Adversarial Networks. In this paper, we propose an end-to-end latent fingerprint synthetic approach based on the improved denoising diffusion probabilistic model. The proposed approach could simultaneously generate latent, rolled, and plain fingerprints of high visual realism. Several primary degradation factors, such as various background textures, limited area of ridge patterns, and structural noise, can be directly generated without any postprocessing, unlike existing methods. We conduct NFIQ2 and perceptual analysis in the experiments to evaluate the proposed approach. The results indicate that the quality and visual realism of the proposed synthetic fingerprints is similar to the natural ones, demonstrating the effectiveness of our approach.

The majority of contemporary fingerprint synthesis is based on the Generative Adversarial Network (GAN). Recently, the Denoising Diffusion Probabilistic Model (DDPM) has been demonstrated to be more effective than GAN in numerous scenarios, particularly in terms of diversity and fidelity. This research develops a model based on the enhanced DDPM for fingerprint generation. Specifically, the image is decomposed into sub-images of varying frequency sub-bands through the use of a wavelet packet transform (WPT). This method enables DDPM to operate at a more local and detailed level, thereby accurately obtaining the characteristics of the data. Furthermore, a polynomial noise schedule has been designed to replace the linear noise strategy, which can result in a smoother noise addition process. Experiments based on multiple metrics on the datasets SOCOFing and NIST4 demonstrate that the proposed model is superior to existing models.

Optical coherence tomography (OCT) is renowned for its high resolution and ability to capture the 3D structure of fingertip skin, significantly enhancing the anticounterfeiting capabilities of fingerprint recognition systems. However, the scarcity of OCT fingerprint datasets, exacerbated by data collection challenges and privacy concerns, poses a major hurdle for practical implementation. We propose a novel conditional diffusion model that generates highly realistic OCT fingerprints from segmentation masks, marking the first attempt to synthesize such images. By modifying the noise model in the diffusion process to account for speckle noise, our method achieves accurate noise simulation and effective removal, resulting in clearer detail feature generation. Subjective evaluations and multiple objective metrics confirm the superior visual quality and diversity of the generated images. By incorporating these images into training datasets for presentation attack detection (PAD) and fingerprint layer segmentation tasks, our method achieves pixel distributions highly consistent with bona fide fingerprints and learns detailed skin structures through segmentation mask guidance. These results highlight the potential of our approach to enhance the performance of OCT fingerprints in practical applications.

Four traditional GAN-based I2I translation techniques for unpaired data have been employed for the synthesis of biometric finger vein presentation attack instrument (PAI) samples in earlier work (three public presentation attack datasets have been considered). These synthetic samples have been used to train presentation attack detectors (PAD). Here we extend this work by using more recent image synthesis techniques to generate the required attack sample data, i.e. StyleSwin and DDPM diffusion. We aim to assess if these more recent techniques are able to outperform the classical GAN techniques when used as training data (using DenseSIFT feature sets in their PAD classifier). Our analysis reveals that, contrasting to expectations, PAD accuracy is on par with the traditional GAN-based synthesis techniques under the restrictive conditions of our evaluation (in particular the severely limited size of available training data).