FACIAL SOFT BIOMETRICS - Library of Ph.D. Theses | EURASIP

22 2. SOFT BIOMETRICS: CHARACTERISTICS, ADVANTAGES AND RELATED WORK2.3.1 Facial soft biometric algorithmsFormer work on soft biometrics has been performed predominantly with the aim of preprocessing.In face recognition for person identification, for instance, beard detection and removal servesan improvement of recognition results, disregarding the information of the presence of beard.Color based facial soft biometrics: The color based facial soft biometric traits (eye, skin,and hair color) are the most obvious facial identifiers, mentioned primarily by humans, whenportraying unknown individuals. Challenges for skin classification are on the one hand the lowspread of different skin colors in color space, and as a consequence, on the other hand the highillumination dependance of classification. Latter is described in various skin locus papers, forexample in [HPM02].Hair color is detected by similar techniques like skin color and often researched along, buthas more broadly scattered color categories. In [SBYL02] a method for human head detectionbased on hair–color is proposed through the use of Gaussian mixture density models describingthe distribution of hair color. In [GHJW00] the fuzzy theory is used to detect faces in color images,where two fuzzy models describe the skin color and hair color, respectively.Eye color classification, unlike the other color based facial soft biometrics is a relatively newresearch topic. Few publications offer insight, see [BRM + 06] and section 7.2, probably due tothe fact that 90% of humans possess brown eyes. An advantage of eye color categorization is theavailability of all necessary information in images used for iris pattern analysis, in other words iriscolor is a free side effect. Work on fusion between iris texture and color can be found in [ZESH04],where the authors fuse iris and iris color with fingerprint and provide performance improvementin respect with the unimodal systems. In [PS08] iris color is used to successfully support an irisindexing method.Beard and Moustache detection: Presence of beard and moustache do not appear in the literatureas an identification trait, but rather as an obstacle for face recognition, which is why theirremoval is performed as a preprocessing step. As an example, in [KM06] a beard removal algorithmis shown using the concept of structural similarity and coordinate transformations.Age: Age plays an important role for long time employable systems based on face or bodyand is a challenging and relatively new field. An interesting study on face changes over time canbe found in [PSA + 07], which spans a biometric, forensic, and anthropologic review, and furtherdiscusses work on synthesizing images of aged faces. In [WM07] the authors distinguish childrenfrom adults based on the face/iris size ratio. Viola–Jones face detection technique [VJ01b] isused, followed by an iterative Canny edge detection and a modified circular Hough transformfor iris measuring, with good results. In [NBS + 08] the authors observe facial skin regions ofCaucasian women and build partial least square regression models to predict the chronologicaland the perceived age. They find out that the eye area and the skin color uniformity are the mainattributes related to perceived age.Gender: Gender perception and recognition has been much researched already in social andcognitive psychology work in the context of face recognition. From image processing point ofview, the topic offers as well many of approaches. A basic approach of understanding simplemetrology in connection to gender is offered in [CCP + 11]. The latest efforts employ a selectionof fused biometric traits to deduce gender information. For example in [CSM07] gait energy imagesand facial features are fused and classified by support vector machines. In [YHP11] contrastand local binary patterns are fused. Another approach in [ST06] proposes a combined gender andexpression recognition system by modeling the face using an Active Appearance Model, featureextraction and finally linear, polynomial and radial based function based support vector machinesfor classification. The work in [BR06] proposes using adaboost on several weak classifiers, ap-