Automatic Segmentation of Eyebrows for Biometric Recognition Using Modified Level Set

Recent research results have suggested the effectiveness of eyebrow features for the tasks of biometric recognition and soft biometric (gender) classification. To date, the problem of eyebrow segmentation has been investigated by few. In this paper, a novel method of eyebrow segmentation based upon a modified level set method is presented. The algorithm first employs adaptive image enhancement methods to detect the dark regions and eye from facial images. Then an estimated region is obtained for the initialization of a modified level set method. The output of the algorithm is a mask of the detected eyebrow. The effectiveness of the proposed approach is demonstrated utilizing a large, challenging set of images from the Facial Recognition Technology (FERET) Database which contain instances of various skin colors, illumination, and partial occlusions.