Title:  3-D Reconstruction of Faces
 
 Project Members:
 Chuoling Chang <chuoling@leland>
 Zhifei Wen <wenzf@leland>
 Edward Li <edwardo@leland>
 
   In this project, we plan to extract feature points on two stereo 2-D images (calibrated or uncalibrated) of a human face.  Begin with the two sets of feature points, we can estimate the disparity between the stereo images and then ascertain the third (depth) dimension information. Finally, we can reconstruct a 3-D model of the original face.
   The goal of the feature point extraction is to find the control points within a 2-D image of a face.  The process to attain this goal will involve face detection, facial feature (eyes, nose, mouth) identification, and then control point localization.
   For the stereo imaging portion we hope to find the third dimension of the entire face by disparity estimation of the two stereo images. At this point, since it is not known how difficult this will be, the secondary plan is to find the third dimension for just the control points.
   Once we have the control points and the third dimension of the face determined, we can generate a 3-D model of the face with the feature points labeled. This 3-D model can be used in many applications such as video communication, computer animation, and novel view generation to name a few.
   However, if we are only able to calculate the third dimension of just the control points, then the scaled back goal will be to fit these control points to simplified generic 3-D face models.  An application of this goal is facial expression transition.  That is, changing a frown to a smile involves control point manipulation. These 3D control points are also useful for facial expression tracking.
   Fundamentally, this project combines two aspects of digital image processing- facial feature extraction and disparity estimation for stereo images.