POSTER: 3D-Face model Reconstruction Utilizing Facial Shape Database from Multiple Uncalibrated Cameras



In recent years, a surveillance camera has come to be attached in various places from a rise of the consciousness to security. Since the surveillance cameras are installed in variety of place, it is possible to take a picture of the same person from multiple uncalibrated cameras though it is asynchronous. In this article, we propose a method for reconstructing a face shape from multiple-view images taken with non -synchronous multiple cameras. In this method, we do not directly reconstruct the shape, but estimate a small number of parameters which represent the face shape. The parameter space is constructed with Principal Component Analysis of database of a large number of anatomical face shapes collected for different people. From the input multiple view images, the region of the face and three feature points on the face are manually extracted. Then the facial pose is estimated by optimizing the evaluation based on the silhouette shape, appearance, and the position of the feature points. According to the facial pose, the parameters representing the facial shape are also estimated by optimizing the same evaluation function. Those optimizing procedures are repeated for obtaining the facial shape for the object face captured with the non-synchronous multiple cameras. The experimental results demonstrate the effectiveness of the proposed method. Since the database used in this paper consists of anatomically aligned shape data, we can obtain anatomical shape of the face, which is suitable to represent the identity of each person.

Download full-text


Available from: Masaaki Mochimaru
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We propose a method that can reconstruct both a 3D facial shape and camera poses from freehand multi-viewpoint snapshots. This method is based on Active Shape Model (ASM) using a facial shape database. Most ASM methods require an image in which the camera pose is known, but our method does not require this information. First, we choose an initial shape by selecting the model from the database which is most suitable to input images. Then, we improve the model by morphing it to fit the input images. Next, we estimate the camera poses using the morphed model. Finally we repeat the process, improving both the facial shape and the camera poses until the error between the input images and the computed result is minimized. Through experimentation, we show that our method reconstructs the facial shape within 3.5 mm of the ground truth.
    Full-text · Conference Paper · Nov 2009