Face+Recognition

=Facial Recognition in Web 2.0= toc

How it Works
First, a facial recognition system scans either a digital image or video to detect faces and seperate them from the background of the image. Because faces are three dimensional the software then has to align them facing forward. This is done in order for the software to then identify specific facial markers and use a biometric or algorithmic system to analize them.

For more information about how facial recognition works see "Biometrics: A Look at Facial Recognition" a publication by the RAND Corporation (Woodward, 2003).

Then the software compares the extracted markers to previously identified images stored in a facial database, or in real time to live video. (Li, 2005, p. 1). The following graphic from Li describes the face recognition process.



A Brief History
Facial recognition has various applications. Upon its inception it was typically used for security purposes by government agencies or large private firms. For example, airports, border portrol, and many federal buildings utilize facial recognition software to improve security measures (Bonsor, 2001). The software currently has a variety of applications including law enforcement, banking and gameing (Huang, 2005). See Table 16.2 from //Handbook of Face Recognition// for more applications.



Facial Recognition and Libraries
Today many photo editing software packages come with embedded recognition systems such as Google's Picassa and Apple's iPhoto software. When Google launched Picassa 3.0 it included name tags for the first time. The "tags use advanced technology to automatically group similar faces together" (Horowitz, 2008).

With the increase in digital image collections at libraries this software can be invaluable for library and archival purposes. Because of the tagging features associated with the face recognition software a user can tag known individuals in a collection and let the software then scan the rest of the collection of images for matches. With most software packages the user then has the option of confirming or denying a match. This semi-automated process helps save archives and archivists time and improves access.