2-D Face Recognition
 | Watch the movie (920 kB) |
The goal of this project is to have a computer recognize a person from an image of his or her face. There are many applications for face recognition. Some examples are: access control, summarizing surveillance video, browsing image and video databases and user-interfaces. The face recognition problem is broken into two important steps. The first is aligning the face to a standard position, size and rotation. This is done by finding a number of facial feature points (such as the corners of the eyes, the tip of the nose, etc) and mapping those points to standard positions. The next step is to compare two aligned faces to get a similarity score. As the major improvement in the past year, we reduced the memory size by 75% and the error rate by 50%.
Background & Objective: Our objective is to develop a state of the art face recognition system. We have been mostly focused on an access control scenario in which the user must cooperate to gain access to a secure room or building. In this scenario lighting can be controlled to a large extent to insure good image quality. We are now beginning to concentrate on less controlled scenarios such as walk-through face recognition or surveillance scenarios in which pose and lighting are much less controlled.
Technical Discussion: The face detection and alignment step is done using the Viola-Jones detection framework developed at MERL. This framework yields very fast and accurate detectors for finding faces and facial feature points. The next step of comparing two aligned faces is currently done using quantized features that are simple Haar-like wavelets. These features are selected using the AdaBoost learning algorithm to best separate pairs of face images of different people from pairs of face images of the same person. After the recognition classifier is trained on a large set of example faces, new pairs of faces that have never been seen before can be compared.
Contacts:
Michael Jones
Jay Thornton
| Technical Reports: | |
| Face Recognition in Subspaces | |
| Face Recognition Using Boosted Local Features | |
| Bayesian Face Recognition | |
Technology Areas:
Imaging
Computer Vision
Modification Date: September 17, 2007
