ISO IEC 19794-4:2011 pdf – Information technology一Biometric data interchange formats一 Part 4: Finger image data.
B.3 Requirements and test procedures for optical fingerprint scanners
B.3.1 Introduction
This annex details requirements and testing procedures for high quality optical fingerprint scanners.
B.3.2 Testing prerequisites
B.3.2.1 Requirements on the testing laboratory
All measurements have to be performed WTthin a completely darkened optical laboratory without the influence of external light sources. The insensitivity of the scanner to external stray tight is not subject of the tests to be performed. For some of the measurements it is necessary to extract light which is emitted by the scanner va prisms: this strongly enhances the sensitivity of the scanner with respect to false light. An exception here is the recording of fingerprints to test the gray scale range. For this test the normal room Illumination has to be switched on. to ensure normal environment conditions similar to me tical usage of the device, While carrying out the measurements it has to be ensured that the optical surface of the fingerprint recording area has to be cleaned. For performing the tests on the scanner the test lab uses the following test tools:
• suitable software for data evaluation (Clause B.3.2.3)
• spreadsheet software
• suitable test targets (Clause B32.4)
The personal of the test lab has to have fundamental knowledge on the test of optical systemsiinstruments. especially on the test of fingerprint scanners.
B.3.2.2 Requirements on the test object
For the test of the fingerprint scanner the manufacturer has to state the exact optical principle of the scanner. including necessary drawings (or pictures, tables). An image capture area of at least 16 mm x 20 mm is required.
The fingerprint scanner to be tested has to be hilly functional, Adaptive or dynamic adjustment. calibration algorithms or spoof detection mechanisms inside the scanner or the scanner software (on the PC), which may include filters, compensation, optimization, dynamic contrast adjustment. have to be disabled dunng the test. For this purpose the manufacturer may have to provide an adapted software for the scanner in which such software parlslalgorithms are deactivated. The software has to operate with constant parameter settings during the test. Only for testing the gray scale range of fingerprint images dynamic algorithms which will be used in customer applications are allowed.
B.3.2.3 Requirements on the evaluation software
The software to evaluate the fingerprint digital image data has to compute image quality based on the two- dimensional spatial frequency power spectrum of the fingerprint digital Image. The power spectrum, which Is the square of the magnitude of the image’s Fourier transform, contains information on the sharpness, contrast. and detail rendition of the image. These are components of visual image quality. Within the software, the power spectrum is normalized by image contrast, average gray level (brightness), and image size; a visual response function filter is applied, and the pixels per inch (ppi) spatial sampling rate scale of the fingerprint image is taken into account. The fundamental oulput Is a singlenumber image quality value which Is the sum of the fihered, scaled, weighted power spectrum values. The power spectrum normalizations allow valid comparisons between disparate fingerprint images. The software has to work as described in the following list:
• The sofiware shall have the digital fingerprint image as input.
• It shall define a square window width of about 60% of fingerprint image width.
• It shall locate the left right and bottom / top edges of the fingerprint.
It shall define a set of overlapping windows covering the entre fingerprint area.
It shall exclude very dense and very low structure areas within the fingerprint from further evaluation.
●It shall compute the 2-D power spectrum of each window and IFFTI2.
●It shall be normalized by total energy and window size.
It shall apply a Human Visual System (HVS) filter (inclusion of such a fiter makes the final quality
values more closely correspond to human observer assessments of relative quality).
●It shall use an initial image quality value per window, i.e. the 2-D normalized, filtered power spectrum values at non-zero frequencles are summed, resulting in a single quality number for the given sub image.
It shall identify the window with the highest image quality.
It shall convert the image quality to the de normalized image quality, that means it has to scale the fingerprint
image to them range [0,100]，where 0 is the worst qualty, 100 is the best qualty.
The image quality overestimates dark areas within the fingerprint images and underestimates bright areas. This effect shall be compensated by multiplying the image quality value with the square of the average gray values.
It shall check for special cases (very high contrast or very light, structured image) and adjust the image quality accordingly.
It shall scale by ppi and normalize the image quality to the range [0,100].
B.3.2.4 Demands on the test targets
B.3.2.4.1
Test targets ftor optical fingerprint scanner working on the principle of frustrated total
internal reflection in the bright field.