Correction of Human Face 3D Model Using 2D Images
Maksims Alekseičevs, Aleksandrs Glazs · 2009
Introduction. Currently, the processing and reconstruction of human face 3D models is used in various areas, for example, in medicine, criminalistics, etc. For instance, in medicine the 3D models of faces can be used for medical diagnostics, plastic surgery and creation of face prosthetics. In criminalistics, an actual task is the identification of personality. Among the different methods of such identification, the most convenient one is the method of face identification, because in this case the contact between the person and a measuring device is not necessary. There are many methods of identification by photos (2D) [3,5,7,8,9,10,12,15,17], but there are not always enough images to successfully identify a person. This is why a new task emerges – using a 3D model of a person’s head to obtain new different images of the person. This leads to the necessity of adapting a generic 3D model of a human head to the specific person. There are different methods of creating a 3D model, based one photo [6,11,13,16] or two photos [1,2,4,14]. This work is dedicated to solving the task of creating a human face 3D model, which is adaptable to a specific person. Problem statement and solution algorithm. As an initial 3D model of human head, a model created in a program FaceGen Modeller was selected. The model of the head contained approximately 8200 vertices or control points (Fig. 1).