U.S. flag

An official website of the United States government, Department of Justice.

NCJRS Virtual Library

The Virtual Library houses over 235,000 criminal justice resources, including all known OJP works.
Click here to search the NCJRS Virtual Library

Probability of False Positive with an Innocent Image Processing Routine

NCJ Number
224401
Journal
Journal of Forensic Identification Volume: 58 Issue: 5 Dated: September/October 2008 Pages: 551-561
Author(s)
Fang Li
Date Published
September 2008
Length
11 pages
Annotation
This study determined the probability that an image processing routine would result in a change to an image used in a fingerprint comparison, such that a false positive conclusion would result.
Abstract
The probability that image processing would innocently result in a change that falsely includes a suspect is less than 1 part in 24 raised to the 100th power (i.e., less than the probability of selecting one specific atom from all those in the universe while blindfolded). The objective of this analysis was to determine the probability than an “innocent” (no intention to change the image) image processing step would create a feature in an image that would change the image to the point of creating a false match. The FBI’s Scientific Working Group on Imaging Technology has included in its guidelines a list of the basic tools that work on the image as a whole or large portion of it. These tools can change certain parts of an image, but they also change all other parts of the image to some degree and with the same properties. They are not capable of altering only a single pixel or a selected block and bringing it to specific levels. If the tools put in a special feature in one location, that will be an accident of the specifics of that location, and there are thousands of other such locations that will also be altered in much the same way. Thus, although it may be theorized that a feature might be added to falsely construct a match, thousands of other changes would inevitably result at the same time, thus producing an exclusion. The statistical analysis is described. 3 figures and 2 references