Automatic tracking of eyes is a challenging task, with numerous applications in
biometrics, security, intelligent human-computer interfaces and driver alertness
detection systems. Eyes detection and localization, therefore, are one of the first steps in
approaching such problems. This project describes a method to detect human pupils in
color images using “separability filter”. The technique, first, detects and extracts human
face region using normalized rg color model. Then, from the detected face region,
intensity valleys such as nostrils, eyes, eyebrows and mouth are extracted. These are the
candidates for pupils. Within these candidates, feature points detection method is
applied to detect the possible positions of pupils. Finally, the pupils are extracted as
circle regions by using a filter known as separability filter. Experiments are done on
FERET Color Database which have been categorized into three face types; normal, with
spectacles and closed eyes. Results reveal that separability filter can best detect human
pupils from a normal face type in which an averaged success rate of about 80% is
achieved.