Latent Fingerprint Enhancement Using Generative Adversarial Networks

Joshi, Indu; Anand, Adithya; Vatsa, Mayank; Singh, Richa; Roy, Sumantra Dutta; Kalra, Prem Kumar

doi:10.1109/wacv.2019.00100

Cited by 37 publications

(26 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The integration of Generative Adversarial Network (GAN) for latent fingerprint quality enhancement was discussed by researchers [Joshi et al (2019)] to upgrade the low quality ridges and predict the ridge information. The proposed algorithm comprising two networks: a latent enhancer network and an enhanced fingerprint discriminator network.…”

Section: Latent Fingerprint Image Enhancement Techniquesmentioning

confidence: 99%

Latent Fingerprint Image Enhancement Based on Optimized Bent Identity Based Convolutional Neural Network

Chaudhary¹,

Dimri²

2021

INDJCSE

View full text Add to dashboard Cite

Fingerprints are unique biometric systems (BSs) in which none of the human possesses similar fingerprint structures. It is one of the most significant biometric processes used in the identification of criminals. Latent fingerprints or latents are generated mainly by the finger sweat or oil deposits which is left by the suspects unintentionally. The impressions of latents are blurred or smudgy in nature and not viewed by naked eye. These fingerprints are of low quality, corrupted by noise, degraded by technological factors and exhibit minor details. Latents display consistent structural info when observed as an image. Image Enhancement is necessary in latents, to transform the latent (noisy) image into fine-quality (enhanced) image. In this work, a new image enhancement approach named BI-CNN (Bent Identity-Convolution Neural Network) with Spatial Pyramid Max Pooling (SPMP) model optimized using TSOA (Tunicate Swarm Optimization Algorithm) is presented to produce an enhanced latent at the output. This procedure involves the integration of ROI (Region Of Interest) Estimation, Anisotropic Gaussian Filter (AGF) based Pre-filtering, Fingerprint alignment using Sobel Filter, Intrinsic Feature patch extraction using Optimized BI-CNN, GAT (Graph Attention) network based Similarity Estimation followed by image reconstruction and feedback module. The implementation tool used in this work is PYTHON platform. The proposed optimized BI-CNN framework tested on dual public datasets namely IIITD-latent finger print and IIITD-MOLF have shown enhanced outcomes. Thus, the IIITD -latent fingerprint database obtained 83.33% on Rank-10 accuracy and 39.33% on Rank-25 accuracy.

show abstract

Section: Latent Fingerprint Image Enhancement Techniquesmentioning

confidence: 99%

Latent Fingerprint Image Enhancement Based on Optimized Bent Identity Based Convolutional Neural Network

Chaudhary¹,

Dimri²

2021

INDJCSE

View full text Add to dashboard Cite

show abstract

“…Qian et al [35] propose DenseUnet while Wong and Lai [50] and Li et al [31] propose multitasking auto-encoder explicitly utilizing orientation field information. Joshi et al [23] propose a generative adversarial network (FP-E-GAN) for fingerprint enhancement. A detailed survey on fingerprint enhancement algorithms is conducted by Schuch et al [37].…”

Section: B Fingerprint Enhancementmentioning

confidence: 99%

“…DATABASES To evaluate the effectiveness of of the proposed work, a wide range of challenging fingerprint databases in the public domain are used to conduct the experimental analysis. These databases are briefly described below: For fingerprint enhancement, two state-of-the-art fingerprint enhancement models: DeConvNet [36] and FP-E-GAN [23] are modified to model data uncertainty. The resulting architectures are named DU-DeConvNet and DU-GAN respectively.…”

Section: B Classification Based Modelsmentioning

confidence: 99%

“…The resulting architectures are named DU-DeConvNet and DU-GAN respectively. Training is performed on synthetic dataset as suggested in [23] while testing us conducted on the Rural Indian Fingerprint Database [34]. Both of these baseline models have regression based loss function.…”

Section: B Classification Based Modelsmentioning

confidence: 99%

“…To establish the effectiveness of modelling data uncertainty in fingerprint preprocessing, in general, we also demonstrate its impact in fingerprint enhancement. In this direction, we take two state-of-the-art fingerprint enhancement models DeConvNet [36] and FP-E-GAN [23] and modify them to DU-DeConvNet and DU-GAN to obtain data uncertainty from these baseline architectures. Table V and Figure 6 (a) demonstrate the fact that the fingerprint quality scores are improved for both the baseline enhancement models after modifying them to predict data uncertainty.…”

Section: Generalization Abilitymentioning

confidence: 99%

See 2 more Smart Citations

Data Uncertainty Guided Noise-aware Preprocessing Of Fingerprints

Joshi¹,

Utkarsh²,

Kothari³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

The effectiveness of fingerprint-based authentication systems on good quality fingerprints is established long back. However, the performance of standard fingerprint matching systems on noisy and poor quality fingerprints is far from satisfactory. Towards this, we propose a data uncertainty-based framework which enables the state-of-the-art fingerprint preprocessing models to quantify noise present in the input image and identify fingerprint regions with background noise and poor ridge clarity. Quantification of noise helps the model two folds: firstly, it makes the objective function adaptive to the noise in a particular input fingerprint and consequently, helps to achieve robust performance on noisy and distorted fingerprint regions. Secondly, it provides a noise variance map which indicates noisy pixels in the input fingerprint image. The predicted noise variance map enables the end-users to understand erroneous predictions due to noise present in the input image. Extensive experimental evaluation on 13 publicly available fingerprint databases, across different architectural choices and two fingerprint processing tasks demonstrate effectiveness of the proposed framework.

show abstract