National Institute of Standards and Technology (NIST) Fingerprint Image Quality (NFIQ) is software that links image quality of optical and ink plain impression 500 pixel per inch fingerprints to operational recognition performance. This allows quality values to be tightly defined and then numerically calibrated, which in turn allows for the standardization needed to support a worldwide deployment of fingerprint sensors with universally-interpretable image qualities. NFIQ 2 quality features are formally standardized as part of ISO/IEC 29794-4. This repository serves as a formally-recognized reference implementation of the 2024 international standard.
Pre-built versions of the NFIQ 2 library and standalone executable for many platforms are available to download on the GitHub Releases page.
In 2004, NIST developed the first open source and publicly available fingerprint quality assessment tool, NFIQ. NFIQ 2 is a revision of that tool. Advances in fingerprint technology since 2004 necessitated an update to NFIQ. As such, development of NFIQ 2 was initiated in 2011 as collaboration between NIST and Germany's Federal Office for Information Security (BSI) and Federal Criminal Police Office (BKA), as well as research and development entities MITRE, Fraunhofer IGD, Hochschule Darmstadt (h_da), and Secunet. Subsequent development efforts and improvements have been supported by experts from ISO/IEC JTC 1/Subcommittee 37 and the community.
NFIQ 2 provides a higher resolution quality score in the range of [0-100], adhering to the international biometric sample quality standard ISO/IEC 29794-1:2024 (as opposed to the original NFIQ's 5-1), lower computation complexity, and support for quality assessment on mobile platforms.
NFIQ 2 is formally recognized as a reference implementation of the normative metrics presented in ISO/IEC 29794-4:2017 and was updated in sync as version 2.3 with the second edition revision effort, ISO/IEC 29794-4:2024.
Operationally, NFIQ has increased the reliability, accuracy, and interoperability of fingerprint recognition systems by identifying the samples that are likely to cause recognition failure.
If you would like more information, please read the NFIQ 2 Report and ISO/IEC 29794-4:2024.
Building the NFIQ 2 library requires the following dependencies, included in this repository as git submodules:
If building the standalone command-line executable, additional dependencies are required, included in this repository as git submodules:
- Biometric Evaluation Framework (public domain license)
- Requires other non-bundled dependencies, please see the README.
- NIST Fingerprint Image Resampler (public domain license)
- Requires OpenCV, which is required by NFIQ 2 library.
Important
Unless you are actively developing code for NFIQ 2, we highly suggest you download from Releases instead of attempting to compile.
Note
You must recursively clone the repository to retrieve git submodules (i.e., do not use the GitHub ZIP file download).
git clone --recursive https://github.com/usnistgov/NFIQ2.git
cd NFIQ2
mkdir build
cd build
cmake .. -DBUILD_NFIQ2_CLI=OFF
cmake --build .Important
Unless you are actively developing code for NFIQ 2, we highly suggest you download from Releases instead of attempting to compile.
Note
You must recursively clone the repository to retrieve git submodules (i.e., do not use the GitHub ZIP file download).
git clone --recursive https://github.com/usnistgov/NFIQ2.git
cd NFIQ2
mkdir build
cd build
cmake ..
cmake --build .-
Standard CMake arguments are interpreted.
- On Windows, change architectures with
-A x64or-A Win32 - On macOS, use the environment variable
CMAKE_OSX_ARCHITECTURESto build a universal binary. Ensure that all dependencies are built universally as well. - Change generators with
-G - Change build types with
-DCMAKE_CONFIGURATION_TYPESor-DCMAKE_BUILD_TYPE
- On Windows, change architectures with
-
Dependencies for
libbiomevalmust be satisfied.- On Windows with Visual Studio, this is done with
vcpkg, which will require passing
the vcpkg
CMAKE_TOOLCHAIN_FILEandVCPKG_TARGET_TRIPLEToptions to CMake.- For example, a 64-bit Release-only build with the default Visual
Studio generator might look like:
cmake .. -DCMAKE_TOOLCHAIN_FILE=C:\vcpkg\scripts\buildsystems\vcpkg.cmake -DVCPKG_TARGET_TRIPLET=x64-windows-static -DCMAKE_CONFIGURATION_TYPES=Release -A x64 cmake --build . --config Release
- For example, a 64-bit Release-only build with the default Visual
Studio generator might look like:
- If building a macOS universal binary, be sure all dependencies are
universal as well. If installed via MacPorts, ensure the
+universalvariants are used.
- On Windows with Visual Studio, this is done with
vcpkg, which will require passing
the vcpkg
Builds for other OS can typically find dependencies on the system without intervention.
Originally, all major versions of OpenCV were supported by NFIQ 2. Due to the limited testing resources as well as slight differences in results between versions, NIST has chosen to rely on the latest release of OpenCV 4 as of this writing. Using a different version of OpenCV may result in unstable NFIQ 2 scores and is not supported. Future updates to OpenCV versions should run the conformance test and larger sequestered tests without differences.
- macOS:
- Xcode 10 and later does not support 32-bit applications. In order to build NFIQ 2 for 32-bit macOS, use Xcode 9.4.x.
The CMake builds supports the following options:
BUILD_NFIQ2_CLI(default:ON)- Whether or not to build the standalone command-line executable.
EMBED_RANDOM_FOREST_PARAMETERS(default:OFF)- Whether or not to embed random forest parameters into the library.
EMBEDDED_RANDOM_FOREST_PARAMETER_FCT(default:0)- Friction ridge capture technology code for embedded random forest
parameters. Only valid if
EMBED_RANDOM_FOREST_PARAMETERSisON.
- Friction ridge capture technology code for embedded random forest
parameters. Only valid if
If you found a bug and can provide steps to reliably reproduce it, or if you have a feature request, please open an issue. Other questions may be addressed to the NIST project maintainers.
NFIQ is released in the public domain. See the LICENSE for details.