Face recognition and using ratios of face features in gender identification
Abstract
In this paper, we have developed a system of algorithms for human face identification and for gender classification. The DRLSE level set method is used for identifying the face location, in which we propose to use a reinitialization step to accelerate the speed of finding the face contours. Gabor wavelet transformation is also used to extract the eye and eyebrow regions of a face, from which a set of triplet parameters are created as ratio values in term of the eye and eyebrow features. A three dimensional linear discrimination algorithm is applied to this set of triplet parameters. This gender identification method takes advantage of the invariant ratio of feature distances to build a criterion that is robust and avoids potential problems caused by the change of the field of view (FOV). The criterion is further applied to a set of testing face images to identify the gender of each individual human face, and improved accuracy rate is achieved.