Facial feature representation with directional ternary pattern (DTP): Application to gender classification

Automatic gender classification is an interesting and challenging task that impacts important applications in biometrics, security, surveillance, and human-computer interaction systems. This paper presents an effective facial feature descriptor based on the directional ternary pattern (DTP) for gender classification. The DTP operator encodes the texture information of a local neighborhood by labeling the edge response values in all eight directions around a pixel with three different levels. The proposed encoding scheme employs a threshold in order to differentiate between high-textured and smooth face regions, and thus, ensures the generation of ternary micro-patterns consistent with the local texture property. Histograms generated from the DTP encoded face images are used as the facial feature descriptor. Extensive experiments on images collected from the FERET face database show the superiority of the proposed method against some well-known local pattern-based feature descriptors.