Fork of NIST Biometric Image Software for building and packaging on Linux. Upstream sources are public domain, but modifications are Apache-licensed.
Highlights from NBIS are:
- Fingerprint image quality assessment program (trainable)
- Fingerprint minutiae extraction program
- Fingerprint minutiae matching program
- Fingerprint Level-1 classification program (trainable)
The master branch of this repository is maintained as the code for building packages for Debian unstable. Binary releases are available from an Ubuntu PPA at https://launchpad.net/~mchaberski/+archive/ppa.
The changes that were made to prepare the upstream sources for building as a Debian package include (but are not limited to) the following:
- the removal of imaging library source code (in favor of using imaging library packages in the Debian repositories);
- inclusion of static
makerules files that were previously generated by a custom shell script; - removal of obsolete files used in the upstream build system; and
- modifications to source code to resolve runtime data in a manner that
follows conventions on Debian systems (such as reading
architecture-independent runtime data from a system directory like
/usr/share/nbis).
A more lengthy description of the contents of the NBIS distribution is below, adapted from the NBIS home page:
The NIST Biometric Image Software (NBIS) distribution is developed by the National Institute of Standards and Technology (NIST) for the Federal Bureau of Investigation (FBI) and Department of Homeland Security (DHS). The NBIS distribution contains programs for biometrics processing and analysis.
The NBIS utilities included in this distribution fall under these general categories:
- A fingerprint segmentation algorithm, NFSEG, segments a four-finger plain impression into individual fingerprint images or can be used to remove white space from a rolled fingerprint image.
- A neural-network based fingerprint pattern classification system, called PCASYS, automatically categorizes a fingerprint image into the class of arch, left or right loop, scar, tented arch, or whorl. This is an updated system that includes the use of a robust Multi-Layered Perceptron (MLP) neural network.
- A minutiae detector, MINDTCT, automatically locates and records ridge ending and bifurcations in a fingerprint image. This system includes minutiae quality assessment based on local image conditions.
- A minutiae-based fingerprint matching algorithm, BOZORTH3, that accepts as input minutiae locations generated by the MINDTCT algorithm.
- A fingerprint image quality algorithm, NFIQ, analyzes a fingerprint image and assigns a quality value of 1 (highest quality) to 5 (lowest quality) to the image. The ability to retrain the NFIQ weights is provided with the utilities FING2PAT, ZNORMDAT, and ZNORMPAT.
- A reference implementation of the ANSI/NIST-ITL 1-2007 (AN2K) "Data Format for the Interchange of Fingerprint, Facial, Scar Mark & Tattoo (SMT) Information" standard is included. This reference implementation contains a suite of utilities designed to read, write, edit, and manipulate files formatted according to this interchange standard. The utilities support updated and new record types introduced by this latest version of the standard (Record Types 9, 13, 14, & 15).
- A large collection of general-purpose image utilities are also included to support the processing of fingerprint images. The package includes encoders and decoders for the FBI's Wavelet Scalar Quantization (WSQ) format. Utilities are also provided that support color component interleaving, colorspace conversion, and format conversion of legacy files distributed in NIST fingerprint databases.
As a consequence of prioritizing making the core utilities buildable first, the SIVV utility (spectral validation/verification metric for fingerprint images) is not included in packages built from this repository.