Enrollment
Various image sources can be used as input for the enrollment process: e.g. a still image, a video stream, or a connection to a facial image database. Grey scale images are as good as color images.
The image is analyzed to determine the position and size of one or more faces. The eye centers are located. Images including eye positions are taken as Primary Facial Data.
Normalization: Based upon the found face and eye positions, the image is scaled, rotated and finally presented at a fixed size.
Preprocessing: Standard techniques are applied such as histogram equalization, intensity normalization.
Characteristics of the face are extracted and represented as a vector in a high-dimensional space, where the similarity of faces is a distance between two vectors. The vector represents the Secondary Facial Data and is used for fast computational comparisons.
Secondary Facial Data can be stored in a central database or on a token storage device or can be used directly for verification or identification.