Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

UFQA: Utility guided Fingerphoto Quality Assessment

Amol S. Joshi, Ali Dabouei, Jeremy Dawson, Nasser Nasrabadi

TL;DR

This work tackles fingerphoto quality assessment by introducing UFQA, a self-supervised dual-encoder framework that jointly optimizes global quality prediction and regional quality maps to reflect recognition utility. It uses a holistic, ECDF-based labeling strategy with regional quality weighting to generate reliable labels, and trains a pairwise-encoder model whose embeddings are fused via self-attention to predict quality in the 0–100 range. Empirical results across multiple fingerphoto datasets show UFQA outperforms NFIQ2.2 and other image quality metrics in terms of matching utility, with ablations confirming the value of dual encoders and regional quality estimation. The approach enables more reliable fingerphoto-based recognition, addressing the challenges of illumination, perspective distortions, and regional fingerprint distortions in contactless biometrics.

Abstract

Quality assessment of fingerprints captured using digital cameras and smartphones, also called fingerphotos, is a challenging problem in biometric recognition systems. As contactless biometric modalities are gaining more attention, their reliability should also be improved. Many factors, such as illumination, image contrast, camera angle, etc., in fingerphoto acquisition introduce various types of distortion that may render the samples useless. Current quality estimation methods developed for fingerprints collected using contact-based sensors are inadequate for fingerphotos. We propose Utility guided Fingerphoto Quality Assessment (UFQA), a self-supervised dual encoder framework to learn meaningful feature representations to assess fingerphoto quality. A quality prediction model is trained to assess fingerphoto quality with additional supervision of quality maps. The quality metric is a predictor of the utility of fingerphotos in matching scenarios. Therefore, we use a holistic approach by including fingerphoto utility and local quality when labeling the training data. Experimental results verify that our approach performs better than the widely used fingerprint quality metric NFIQ2.2 and state-of-the-art image quality assessment algorithms on multiple publicly available fingerphoto datasets.

UFQA: Utility guided Fingerphoto Quality Assessment

TL;DR

This work tackles fingerphoto quality assessment by introducing UFQA, a self-supervised dual-encoder framework that jointly optimizes global quality prediction and regional quality maps to reflect recognition utility. It uses a holistic, ECDF-based labeling strategy with regional quality weighting to generate reliable labels, and trains a pairwise-encoder model whose embeddings are fused via self-attention to predict quality in the 0–100 range. Empirical results across multiple fingerphoto datasets show UFQA outperforms NFIQ2.2 and other image quality metrics in terms of matching utility, with ablations confirming the value of dual encoders and regional quality estimation. The approach enables more reliable fingerphoto-based recognition, addressing the challenges of illumination, perspective distortions, and regional fingerprint distortions in contactless biometrics.

Abstract

Quality assessment of fingerprints captured using digital cameras and smartphones, also called fingerphotos, is a challenging problem in biometric recognition systems. As contactless biometric modalities are gaining more attention, their reliability should also be improved. Many factors, such as illumination, image contrast, camera angle, etc., in fingerphoto acquisition introduce various types of distortion that may render the samples useless. Current quality estimation methods developed for fingerprints collected using contact-based sensors are inadequate for fingerphotos. We propose Utility guided Fingerphoto Quality Assessment (UFQA), a self-supervised dual encoder framework to learn meaningful feature representations to assess fingerphoto quality. A quality prediction model is trained to assess fingerphoto quality with additional supervision of quality maps. The quality metric is a predictor of the utility of fingerphotos in matching scenarios. Therefore, we use a holistic approach by including fingerphoto utility and local quality when labeling the training data. Experimental results verify that our approach performs better than the widely used fingerprint quality metric NFIQ2.2 and state-of-the-art image quality assessment algorithms on multiple publicly available fingerphoto datasets.
Paper Structure (18 sections, 4 equations, 9 figures, 1 table)

This paper contains 18 sections, 4 equations, 9 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Fingerprint Quality Assessment
  4. Fingerphoto Quality Assessment
  5. Proposed Method
  6. Quality Label Generation
  7. Quality Assessment Model
  8. Feature extraction
  9. Quality prediction
  10. Implementation Details
  11. Experiments
  12. Datasets
  13. Evaluation Protocol
  14. Results
  15. Ablation Study
  16. ...and 3 more sections

Figures (9)

  • Figure 1: Comparison between fingerphoto quality assessment methods. (a) Conventional approach to obtain quality through pre-processing, such as filtering, histogram equalization, etc. and applying NFIQ2. (b) Our proposed UFQA model that incorporates fingerphoto utility and local information to predict quality.
  • Figure 2: ECDF plot with normalized matching scores on the x-axis and CDF on the y-axis. The dots in the plot show the split location for each class.
  • Figure 3: Quality label generation process flow. The labels obtained in the figure are for the probe set. We switch the gallery and probe sets and repeat the process to get the labels for the gallery set.
  • Figure 4: Generated quality labels for different samples. The number in the upper right corner represents the quality label for each image. Zoom in for better view.
  • Figure 5: The architecture of the proposed UFQA model. The input to the network is two fingerphotos, and the output features from the candidate encoder are used to predict the image quality between $0$ and $100$ using the quality prediction network. The feature fusion module is displayed on the right side of the figure.
  • ...and 4 more figures