Algorithms
id3 Semiconductors proposes two fingerprint recognition algorithms:
- an ISO standards compliant template extractor -id3Extract -,
- an ISO compliant fingerprint on card (Match On Card) matcher - id3FMOC -.
Most of fingerprints recognition systems (1 : 1 or 1 : N) are based on a traditional approach: the extraction and the comparison of few unique marks on the fingerprint commonly known as minutiae. These may be defined as fingerprint ridge endings (point where a ridge stops) or bifurcations (point where a ridge is divided in two ridges).
As shown during years of research and worldwide applications, these features are sufficient to design both accurate and highly interoperable systems.
Aware of the strength of minutiae as a biometric identifier, id3 Semiconductors has developed fast, robust and standardized solutions, fully ISO/IEC 19794-2 compliant.
ISO template extractor - id3Extract - :
id3Extract minutiae extractor algorithm works according to a classical two steps basis:
- Image denoising, enhancement and segmentation,
- Minutiae extraction..
Fig. 1 - Main exctraction's steps: (a) original fingerprint image, (b) enhanced image, (c) extracted minutiae.
The highlights of id3Extract are :
A powerful segmentation :
The creation of false minutiae in the background of the finger or in poor quality areas is one of the main issues of the extraction process that results (during comparisons) in the growth of the false non match rate and false match rates. Thanks to a unique approach, the id3Extract algorithm removes theses problematic areas in order to extract a minimum quantity of false minutiae
An high accuracy :
Advanced mathematical methods precisely forsee the orientation of the ridges and the mean space between them. These features allow a high-quality image filtering of the fingerprint, carrying out the extraction of minutiae even in poor quality or low contrast areas.
Sensor independence:
id3Extract algorithm has been developed to work on every kind of images. This particularity ensures an homogenous minutiae extraction from every type of sensor technology, either optical, capacitive or thermal.
Fig. 3 - Sensor independence : (a) optical sensor, (b) capacitive sensor, (c) thermal sensor
On card comparison and embedded matching algorithm - id3FMOC -
In order to reach the interoperability as better as possible with existing systems, The id3 Semiconductors fingerprint on card comparison algorithm id3FMOC has only been developed in the ground of the ISO/IEC 7816-11 et ISO/IEC 19794-2 standards .
Minutia-based comparison is a complex compromise between computation speed and biometric performance. It is even more complex that some fingerprint images undergo rotations or rotations conditioned by the position of the finger and the pressure exerted during a capture. In such conditions, the resulting incomplete and/or distorted minutiae sets need a more robust algorithm to succeed a genuine comparison, and the more robust the comparison algorithm, the higher the computation. Furthermore, algorithms for integration into a smart card or an embedded platform have a very strong constraints on the amout of calculations, leading logical performance limitations.
In a drastically limited context, id3FMOC algorithm offers an excellent compromise between deformations, computation time and biometric performance, due to many optimizations and a smart seeking technology.
As demonstrated by NIST ("National Institute of Standards and Technology") in the MINEX II ("Minutiae Interoperability Exchange") report, id3FMOC algorithm not only exceeds the PIV ("Personal Identification Verification") specifications in terms of accuracy but is also the fastest on-card comparison implementation to date. With a tiny memory footprint, it is perfectly suitable for most of the smart card commercially available.