Fast and robust self-training beard/moustache detection and segmentation

Abstract

Facial hair detection and segmentation play an important role in forensic facial analysis. In this paper, we propose a fast, robust, fully automatic and self-training system for beard/moustache detection and segmentation in challenging facial images. In order to overcome the limitations of illumination, facial hair color and near-clear shaving, our facial hair detection self-learns a transformation vector to separate a hair class and a non-hair class from the testing image itself. A feature vector, consisting of Histogram of Gabor (HoG) and Histogram of Oriented Gradient of Gabor (HOGG) at different directions and frequencies, is proposed for both beard/moustache detection and segmentation in this paper. A feature-based segmentation is then proposed to segment the beard/moustache from a region on the face that is discovered to contain facial hair. Experimental results have demonstrated the robustness and effectiveness of our proposed system in detecting and segmenting facial hair in images drawn from three entire databases i.e. the Multiple Biometric Grand Challenge (MBGC) still face database, the NIST color Facial Recognition Technology FERET database and a large subset from Pinellas County database.