Resolution & Gray Scale Levels

Effective hardware characteristics for real-world biometrics

The key quality characteristics of a fingerprint sensor besides the dimension of the scan area are the resolution and number of gray scale levels.

Sensor Resolution
The resolution specifies the density of the pixels and is defined by the pixel dimensions in a fingerprint sensor design. It is measured in dpi (dots per inch) just like the resolution of a printer or document scanner. The average human finger ridge width is 0.427 mm and 0.483 mm in female and male adults, respectively. A resolution of 320 dpi will result in 4-5 pixels ridge width and 2-4 pixels valley width - which is enough to reliably detect all minutiae points. 

The recommended resolution for a high quality fingerprint sensor depends entirely on the application. Leading commercial fingerprint area sensors with capacitive or active thermal sensing are available in the range from 363 to 500 dpi. Statistical analysis and scientific research has proven that a higher resolution like 500 dpi used by many vendors has little effect on the accuracy while increasing data size significantly. Some algorithms will when working with 500 dpi sensors actually reduce the resolution to save time. The higher resolution is useful, e.g., when identifying newborn and infants. 

More important than the actual resolution is good image quality and uniformity. The sensor should have only little distortion (divergence from uniform linear projection with increasing distance from the center of the scan area). Distortion is prominent e.g. in cheap optical scanners lacking geometric acuracy.

The NEXT sensor targets 1:1 matching applications and offers a resolution of 385dpi. This is a conservative specification to reliably detect all the unique features (true minutiae) in a fingerprint while keeping the amount of data in a range suitable for Smartphones, tablets, PCs and embedded systems for commercial use.

Why are there Sensors with Different Resolution Specifications? 
Sensor resolution is determined by the sensor design and the nature of the application. 

From a scientific viewpoint, fingerprints include 3 types of features:

level 1: global shape, fingerprint type class
level 2: minutiae (ridge ends, bifurcations and complex friction ridge structures)
level 3: high resolution features (pores (60um), incipient ridges, line shape, ...)

Most commercial fingerprint algorithms rely on minutiae only. They have a long proven track record with statistical analysis over hundreds of millions of fingerprints. Sensors with higher resolution have not been designed for commercial applications.

The following table summarizes the applications and matching technology:

150dpi+
global shape, type classification, fortune telling

320dpi+
minutiae algorithms, access control, financial services, device logon

500dpi+
criminal AFIS, law enforcement, pattern matching

1.000dpi+
forensic sciences, fingerprinting newborn and infants

Gray Scale Levels
The gray scale levels of a fingerprint sensor are the number of gray shades produced for every pixel. 256 levels of gray is the de facto standard supported by most available fingerprint sensors today and results in using one byte per pixel.

Other important aspects for gray scales are the uniformity and dynamic range. Uniformity assesses the gray level difference in a neighboring pixel area while dynamic range measures the gray values actually used from the maximum of typically 256. Requirements depend on the application.