Decremental depth bunch based 3D face recognition from range image

In this paper, a new technique, i.e. decremental depth bunches have been presented where two facial discriminating mechanisms have also been implemented for recognizing the individuals. Notably, based on variations of the depth values, different bunches of face regions (i.e. the small components) are extracted by differentiating the depth information that eventually describes detailed facial surface information. Now, from each bunches, statistical attributes as well as Hough peaks are encountered to initiate two feature vectors for feature-based as well as a holistic mechanism for classification by K-NN and Cosine distance respectively. The proposed mechanism is explicitly dependent on facial depth information that have been accomplished in range face images. Therefore, authors have considered two databases, namely: Frav3D and Bosphorus, that contains laser as well as structured light 3D scanner based procured 3D face images respectively.