Post-graduate theses
Search command : Author="Στεφανίδης"
And Author="Κωνσταντίνος"
Current Record: 4582 of 6549
|
|
| Identifier |
uch.csd.msc//2003simantiris |
| Title |
Παρακολούθηση προσώπων και εκτίμηση της πόζας σε ψηφιακές ακολουθίες εικόνων |
| Alternative Title |
Face tracking and pose estimation in digital image sequences |
| Creator |
Simantiris, George
|
| Abstract |
In this work a face tracking and facial pose estimation methodology is presented. Facial features are tracked and the 3-D motion and geometrical parameters are estimated. Our work is based on the assumption that the face is close to the camera and face motion is smooth and simple. It can be applied in security systems, human-computer interfaces and telecommunication environments. Furthermore, it comprises the basis for more sophisticated image analysis regarding to faces, such as expression analysis-recognition, gaze tracking, lip reading followed by speech recognition and face recognition-authentication. The method is based on points, strategically placed on the face, such that they determine the main facial features, which are eyes, nose and mouth. It consists of two stages~: the estimation of the 3-D motion and geometrical parameters and the 2-D point matching. In the first stage, the selected points are considered to correspond to points onto a planar surface in 3-D space. This modelling technique is simpler and more profitable, in terms of computational cost, comparative to others, presented in the respective research area. The 2-D image positions of the points in two consecutive frames are capable of retrieving the 3-D motion parameters, translation and rotation, as well as the face pose relative to the camera. In the second stage, the points are tracked throughout the image sequence. The motion estimation is used in order to predict the point coordinates in the next frame. This prediction defines a search area, in which an attempt to find the corresponding points takes place, based on an multiple block matching technique in addition to geometrical constraints. To evaluate the proposed methodology, simulations as well as real image sequences have been used. The experimental results satisfy in accuracy towards the cost. Orientation of research towards this direction is encouraged, targeting the development of a face tracking system meeting high standards and real-time specifications.
|
| Issue date |
2003-04-01 |
| Date available |
2003-06-24 |
| Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
|
|
|
Type of Work--Post-graduate theses
|
| Permanent Link |
https://elocus.lib.uoc.gr//dlib/d/9/7/metadata-dlib-2003simantiris.tkl
|
| Views |
436 |
Collections
