Eye detection using color cues and projection functions

We pmpose a heuristic approach for detection ofeyes in close-up images. The experimental images are stereorypiral mug shot faces which can be expected in applications offace recognition systems, say in ATM vestibules. Wepmve the efficacy of our proposed melhod in delection of eyes, both in indoor and outdoor envimnments wilh considerablescale variations and an allowable mlalion in the imageplane. We employa hierarchical searchspacereduclion technique 10 localize posslbie eye areas in the image. Thedistincthumanskin color and low intensityareas of the eye bail are the primary cues used lo locate eye regions. Further on, eye validation is performed using fhe meanand varianceprojection functions. Keywords : Normalized RGB. illumination normalization. connected component analysis, projection functions. eye detection.

• Type: Conference
• Year: 2002
• Where: ICIP
• Authors: Thilak R. Kumar, S. Kumar Raja, A. G. Ramakrishnan
• Keywords: Eye Bail, ICIP 2002, Image Processing, Lo Locate Eye, Posslbie Eye Areas