Products > Biometrics Software > QualityCheck

Automated Fingerprint Identification Systems (AFIS) reject a certain percentage of submitted fingerprint images because they fail to meet image quality criteria.  Poor quality is a problem when it contributes to the inability of a machine or human expert to identify minutiae points or core/delta points in a given fingerprint image. Such failure to extract minutiae points is usually attributed to poor ridge flow, poor contrast and brightness in the image, or partial images. 

Image quality analysis is a critical component of a fingerprint live scan workstation. It provides real-time image quality feedback to an operator, thus helping to reduce the number of poor quality submissions to an AFIS and increasing the value and reliability of a biometric image database.

Aware’s QualityCheck is an advanced fingerprint image quality scoring software library. Designed for systems integrators and government agencies tasked with biometric data collection, QualityCheck uses complex algorithms to assess the quality and usefulness of a fingerprint image, putting a powerful tool in the hands of operators to help improve the overall performance of an AFIS. 

QualityCheck implements a non-AFIS specific measure of finger image quality that is based on the continuity of ridge flow across all regions of a finger image. It is designed to mimic the human visual perception of finger image quality. Quality problems that are apparent to a human expert will be detected by QualityCheck which returns information based on the following questions:

1. Is the image smudged due to movement, improper finger placement, or excess moisture?
   
   
2. Is the image too dark due to excess finger pressure applied to the live scan platen or due to miscalibration of the sensor?
   
   
3. Is the image too light due to inadequate finger pressure applied to the live scan platen or due to miscalibration of the sensor?
   
   
4. Is the image too small?
   
   
5. Does the image lack a core or delta?
   
   
6. Given a large population of similar FBI compliant finger images, how does this image compare?

QualityCheck generates an overall score between 0 and 100, and provides information on areas of the image that exhibit problems. These areas are returned to a software application as arrays of pixel regions or as a color-coded image of the finger, which indicate the specific problems with the finger image. This functionality can improve the ability of an operator to screen bad images.

The finger images shown below, in order from best to worst quality, are samples from field deployed systems. The quality values and color coding information are returned by the Aware QualityCheck functions. The color codes provide quick visual assistance to identify the following gross problems with an image:

Blue = smudged or broken areas
Red = areas that are too dark
Yellow = areas that are too light
Green = areas of good quality

The quality score is measure of ridge flow- a primary indicator of fingerprint image quality. The score range is 0-100. The correlation between the scores and the general quality of an image can be understood by examining the distribution curve shown in the graph below. Each of 17,000 FBI compliant live scanned images (different scanners, impressions, and rolls) are scored and plotted.

Example Classification Thresholds:

0-40 Poor
41-60 Marginal
61-75 Adequate
76-100 Good

Helps identify partial images or images consisting only of fingertips. Image #7 shows an example where the minutiae count is low, and no core or delta was found. Partial finger images can pose a particular problem because they may have good ridge flow, but still do not provide the correct information. Lack of core/delta and low minutiae counts helps to flag those images.

Provides the total count of the green area for each image. This is the part of the image where Minutiae points likely can be extracted from. This number can be used to flag images that are too small.

Provides the total count of the red (too dark), yellow (too light) and blue (broken, smudged) pixels. Images with low ratios of good-to-bad pixels (images #4, 5, 6, and 7) closely correlate with low quality scores.

Describes where the given image falls in the sample distribution shown in the plot below. The value indicates the percent of images from this database of 17,000 FBI-compliant live scanned fingers that exhibited lower scores. 

• Includes C callable library or ActiveX control designed to be integrated into a larger application
• Includes example programs and their 
  source code
• Provides a score indicating quality of finger ridge data
• Manual and auto image cropping functions to remove the finger ridge data from a noisy or large background image
• Provides minutiae counts and number of core/delta found
• Indicates pixel regions that are good, too dark, too light, or that have smudged/broken finger ridges
• Color-coded image-based function returns this same information
• Usable with all matchers and systems
• Statistically reliable results
• Supported by MS Windows, Linux, and Solaris platforms
  

To receive more information about Aware's Biometrics Software Products, please contact us.