At the beginning of the whole process 3D facial surface is captured (example of creating 3D face is at the pictures Fig. 3.4 – Fig. 3.6). There are several different ways how to achieve this task, for example stereo cameras, depth camera, laser, optical or laser scanner, etc.
Only face is needed from whole captured image. Because of that cropping of face is needed. Each face is in the rectangle, which consists of 4 points on the head. The side edges consist from points which position is most left and right. The highest point makes upper edge and the lower edge consist of the lowest point. Then the cropping is based on this rectangle made by these 4 points.
The captured data are subsequently preprocessed using feature extraction algorithms.
The purpose of feature extraction is to extract the compact information from the images that is relevant for distinguishing between the face images of different people and stable in terms of the photometric and geometric variations in the images.
As features can be used facial points (head top, forehead, eyes, chin, nose, mouth, etc.) and distances between these selected points in 3D Euclid space (Fig. 3.7).