面向光学相干层析指纹的条件扩散生成方法

Conditional Diffusion Generation for Optical Coherence Tomography Fingerprints

摘要

光学相干断层扫描技术以其高分辨率和捕捉指尖皮肤三维结构的能力而闻名，能够增强指纹识别系统的防伪能力。然而，与其它生物识别技术相比，数据集的稀缺严重阻碍其广泛应用。由于采集困难且出于隐私考虑不便公开，研究数据生成是应对该挑战的解决方案之一。本文提出一种基于扩散模型的条件生成方法，利用层分割掩码作为先验知识引导生成过程，通过逐步去噪直接在像素空间建模，避免潜在扩散模型的精度损失，从而生成高保真 OCT 指纹图像。实验表明，该方法生成的样本具有逼真的皮肤结构特征，通过对 30 位领域专家进行的主观评价实验证明其生成结果解剖结构清晰、像素分布真实。进一步实验证明，使用生成数据扩充训练集可显著提升多种防伪模型的性能。

Abstract

Optical Coherence Tomography（OCT）is renowned for its high resolution and ability to capture the three-dimensional struc- ture of fingertip skin，significantly enhancing the anti-spoofing capability of fingerprint recognition systems.However，compared to oth- er biometric technologies，the scarcity of datasets severely hinders its widespread application.Due to the difficulty of data collection and privacy concerns that restrict public sharing，synthetic data generation presents a more practical solution to this challenge.This pa- per proposes a conditional generation method based on a diffusion model，which leverages layer segmentation masks as prior knowl- edge to guide the generation process.By employing iterative denoising to directly model the pixel space，our approach avoids the preci- sion loss associated with latent diffusion models，thereby producing high-fidelity OCT fingerprint images.Experiments demonstrate that the generated samples exhibit realistic skin structural features，with subjective evaluations by 30 domain experts confirming their ana- tomically accurate structures and authentic pixel distributions.Further validation shows that augmenting training datasets with synthetic data significantly improves the performance of various anti-spoofing models.

关键词

光学相干断层扫描 / 指纹 / 条件生成 / 伪造攻击检测

Key words

optical coherence tomography / fingerprint / conditional generation / presentation attack detection

参考文献

[1] Cappelli R, Ferrara M, Maltoni D. Minutia cylinder-code:a new representation and matching tcchniquc for fingcrprint rccognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelli- gence, 2010, 32(12):2128-2141.

[2] Das A, Galdi C, Han H, et al. Recent advances in biometric tech- nology for mobile devices[C]// Proceedings of IEEE 9th Interna- tional Conference on Biometrics Theory,Applications and Systems (BTAS),2018:1-11.

[3] Khan T M, Bailey D G, Khan M A U, et al. Efficient hardware im- plementation for fingerprint image enhancement using anisotropic gaussian filter[J]. IEEE Transactions on Image Processing, 2017, 26(5):2116-2126.

[4] Chugh T, Jain A K. Oct fingerprints: resilience to presentation at- tacks[EB/OL]. https://arxiv.org/abs/1908.00102,2019.

[5] Goicoechea Telleria I, Fernandez Saavedra B, Liu Jimenez J, et al. An evaluation of presentation attack detection of fingerprint biomet- ric systems applying iso/iec 30107-3[C]// Proceedings of 4th In- ternational Biometric Performance Testing Conference(IBPC), Gaithersburg, 2016.

[6] Information technology-biometric presentation attack detection- part 1:framework[M]. Geneva: International Standards Organiza- tion, 2016.

[7] Sousedik C, Busch C. Presentation attack detection methods for fin- gerprint recognition systems:a survey[J]. IET Biometrics, 2014, 3 (4):219-233.

[8] Liu F. Liu G, Wang X. High-accurate and robust fingerprint anti- spoofing system using optical coherence tomography[J]. Expert Systems with Applications, 2019, 130(9):31-44.

[9] Bontrager P, Roy A, Togelius J, et al. Deepmasterprints :generating masterprints for dictionary attacks via latent variable evolution[C]// Proceedings of IEEE 9th International Conference on Bio- metrics Theory,Applications and Systems(BTAS),2018:1-9.

[10] Liu F, Zeng W, Li Y, et al. A lightweight and noise-robust method for internal OCT fingerprint reconstruction[J]. IEEE Transactions on Information Forensics and Security, 2024, 19:5492-5505.

[11] Sun H, Wang H, 7hang Y, et al. Zjut-eifd:a synchronously collect- ed external and internal fingerprint database[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 46(4):2267-2284.

[12] Liu F, Shen C, Liu H, et al. A flexible touch-based fingerprint ac- quisition device and a benchmark database using optical coherence tomography[ J]. IEEE Transactions on Instrumentation and Meas- urement, 2020, 69(9):6518-6529.

[13] Zhu Y, Yin X, Hu J. Fingergan:a constrained fingerprint generation scheme for latent fingerprint enhancement[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45 (7):8358-8371.

[14] Engelsma J J, Grosz S A, Jain A K. Printsgan:synthetic tingerprint generator[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(5):6111-6124.

[15] Grosz S A, Jain A. Spoofgan: synthetic fingerprint spoof images[J]. IEEE Transactions on Information Forensics and Security, 2022, 18:730-743.

[16] Sohl Dickstein J N, Weiss E A, Maheswaranathan N, et al. Deep unsupervised learning using nonequilibrium thermodynamics[EB/OL]. https://arxiv.org/abs/1503.03585,2015.

[17] Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models[EB/OL]. https://arxiv.org/abs/2006.11239,2020.

[18] Wu YN, Gao R, Han T, et al. A tale of three probabilistic families: discriminative,descriptive and generative models[EB/OL]. https://arxiv.org/abs/1810.04261,2018.

[19] Kingma D P, Welling M. Auto-encoding variational bayes[EB/OL]. https://arxiv.org/abs//1312.6114,2022.

[20] Goodfellow I J, Pouget Ahadie I, Mirza M, et al. Generative adver- sarial networks[J]. Communications of the ACM, 2014, 63(11): 139-144.

[21] Dinh L, Sohl Dickstein J N, Bengio S. Density estimation using real nvp[EB/OL]. https://arxiv.org/abs/1605.08803,2016.

[22] Kingma D P, Dhariwal P. Glow:generative flow with invertible 1x1 convolutions[EB/OL]. https://arxiv.org/abs/1807.03039,2018.

[23] Dhariwal P, Nichol A. Diffusion models beat GANs on image syn- thesis[EB/OL]. https://arxiv.org/abs/2105.05233,2021.

[24] Kazerouni A, Aghdam E K, Heidari M, et al. Diffusion models in medical imaging:comprehensive survey[J]. Medical Image Analy- sis, 2023, 88:102846.

[25] Han C, Hayashi H, Rundo L, et al. Gan-based synthetic brain mr image generatio[C]// Proceedings of 15th IEEE International Symposium on Biomedical Imaging(ISBI), 2018:734-738.

[26] Yang Q, Yan P, Zhang Y, et al. Low dose ct image denoising using a generative adversarial network with wasserstein distance and per- ceptual loss[J]. IEEE Transactions on Medical Imaging, 2017, 37 (6):1348-1357.

[27] Song J, Meng C, Ermon S. Denoising diffusion implicit models[EB/OL]. https://arxiv.org/abs/2010.02502,2020.

[28] Song Y, Sohl Dickstein J N, Kingma D P, et al. Score-based gener- ative modeling through stochastic differential equations[EB/OL]. https://arxiv.org/abs/2011.13456,2020.

[29] Rombach R, Blattmann A, Lorenz D, et al. High-resolution image synthesis with latent diffusion models[C]// Proceedings of IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2021:10674-10685.

[30] Cappelli R, Maio D, Maltoni D, et al. Synthetic fingerprint-image generation[C]// Proceedings of 15th International Conference on Pattern Recognition(ICPR), 2000:471-474.

[31] Zhao Q, Jain A K, Paulter N G, et al. Fingerprint image synthesis based on statistical feature models[C]//Proceedings of IEEE 5h International Conference on Biometrics: Theory,Applications and Systems(BTAS), 2012:23-30.

[32] Johnson P A, Hua F, Schuckers S. Texture modeling for synthetic fingerprint generation[C]// IEEE Conference on Computer Vision and Pattern Recognition Workshops(CVPRW), 2013:154-159.

[33] Kirchgasser S, Kauba C, Uhl A. Assessment of synthetically genera- ted mated samples from single fingerprint samples instances[C]// Proceedings of IEEE International Workshop on Information Foren- sics and Security( WIFS), 2021:1-6.

[34] Bouzaglo R, Keller Y. Synthesis and reconstruction of fingerprints using generative adversarial networks[EB/OL]. https://arxiv.org/abs/2201.06164,2022.

[35] Bahmani K, Plesh R, Johnson P A, et al. High fidelity fingerprint generation:quality,uniqueness,and privacy[C]// Proceedings of IEEE International Conference on Image Processing(ICIP), 2021: 3018-3022.

[36] Wyzykowski A B V, Segundo M P, de P Lemes R. Level three syn- thetic fingerprint generation[C]// Proceedings of 25th International Conference on Pattern Recognition(ICPR), 2020:9250-9257.

[37] van den Oord A, Vinyals O, Kavukcuoglu K. Neural discrete repre- sentation learning[C]// Proceedings of 31st International Confer- ence on Neural Information Processing Systems( NIPS ), 2017; 6309-6318.

[38] Razavi A, Van den Oord A, Vinyals O. Generating diverse high-fi- delity images with vq-vae-2[C]// Proceedings of 33rd Internation- al Conference on Neural Information Processing Systems (NeurIPS), 2019:14866-14876.

[39] Isola P, Zhu J Y, Zhou T, et al. Image-to-image translation with conditional adversarial networks[C]// Proceedings of IEEE Con- ference on Computer Vision and Pattern Recognition(CVPR), 2017:5967-5976.

[40] Karras T, Aittala M, Laine S, et al. Alias-free generative adversarial networks[C]// Proceedings of the 35th International Conference on Neural Information Processing Systems(NeurIPS), 2021:852-863.

[41] Li B, Xue K, Liu B, et al. BBDM:image-to-image translation with brownian bridge diffusion models[C]// Proceedings of IEEE Con- ference on Computer Vision and Pattern Recognition(CVPR), 2023:1952-1961.

[42] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323:533-536.

[43] Gong D, Liu L, Le V, et al. Memorizing normality to detect anoma- ly:memory-augmented deep autoencoder for unsuperv-ised anoma- ly detection[C]// Proceedings of IEEE International Conference on Computer Vision(ICCV), 2019:1705-1714.

[44] Liu F, Liu H, Zhang W, et al. One-class fingerprint presentation at- tack detection using auto-encoder network[J]. IEEE Transactions on Image Processing, 2021, 30:2394-2407.

[45] Liu Y P, Zuo W, Liang R, et al. Prototype-guided autoencoder for OCT-based fingerprint presentation attack detection[J]. IEEE Transactions on Information Forensics and Security, 2023, 18: 3461-3475.

[46] Zhang W, Liu H, Liu F, et al. A uniform representation model for oct-based fingerprint presentation attack detection and reconstruc- tion[J]. Pattern Recognition, 2024, 145: 109981.

[47] Liu Y P, Zhong Q, Liang R ct al. Layer segmentation of oct finger- prints with an adaptive Gaussian prior guided transformer[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71:1-15.

基金资助

国家白然科学基金项目(62376251)

国家白然科学基金项目(62076220)

浙江省医药卫生科技计划项目(2024 KY881)