Robust detection of facial features by generalized symmetry

Abstract

Locating facial features is crucial for various face recognition schemes. We suggest a robust facial feature detector based on a generalized symmetry interest operator. No special tuning is required af the face occupies 15-60% of the image. The operator was tested on a large face data base with a success rate of over 95%.

Detecting Faces in Impoverished Images

“…Specifically, we were interested in determining whether misorientations along certain axes were particularly disruptive for performance. Vertical bilateral symmetry is often considered an important defining attribute of faces [Reisfeld and Yeshurun, 1992;Thornhill and Gangested, 1993;Sun et al, 1998]. We expected, therefore, that detection performance would be disrupted disproportionately for orientations that destroyed the vertical bilateral symmetry.…”

Image Warping by Radial Basis Functions: Application to Facial Expressions

“…We have recently demonstrated the applicability of such an approach to face recognition. We have developed a system that automatically detects the most important facial features (eyes and mouth) using generalized symmetry 26,5]. We have also shown that normalizing a 2D image of a face using an a ne transformation determined by the location of the eyes and mouth is an e ective step towards face recognition 11].…”

Face recognition technology for law enforcement applications

“…Early work on face recognition was done by Sakai et al in [105] [104]. From their paper it is not apparent how they determine whether a feature is the eye or the nose, i.e.…”