Prof. Xiaoou Tang
Department of Information Engineering
The Chinese University of Hong Kong
Automatic face recognition has great potential in a large array of application areas, including banking and security system access control, video surveillance, mug-shot matching for law enforcement, duplicate ID card detection, and multimedia information retrieval, etc. Face recognition research has attracted great attention in recent years. In this tutorial, we first give a brief introduction to face recognition technology, and then review some basic algorithms including PCA, LDA, Bayesian, and EGM. The focus of the tutorial will be on recently developed new algorithms in face processing research. We will describe a new framework that unifies all the current subspace face recognition algorithms. Under this framework, we can clearly see the advantage and disadvantage of each individual algorithm and discover the inherent relationship among different subspaces as well as their unique contributions to the extraction of discriminating information from the face images.
In addition to photo-based face recognition, we will also explore an interesting new topic, sketch-based face recognition. Automatic retrieval of photos of suspects from police mug-shot database can help the police narrow down potential suspects quickly. However, in most cases, the photo image of a suspect is not available. The best substitute is often a sketch drawing based on the recollection of an eyewitness. Automatically searching through a photo database using a sketch drawing is very useful. It will not only help the police to locate a group potential suspects, but may also help the witness and the artist to modify the sketch drawing of the suspect interactively based on the similar photos retrieved. In this tutorial we will introduce sketch recognition algorithms that can achieve recognition accuracy much higher than human performance.
Another interesting topic we will discuss is face hallucination. In video surveillance, the faces of interest are often in small size because of the large distance between the camera and the objects. Image resolution becomes an important factor affecting face recognition performance. Since many detail facial features are lost in the low-resolution face images, the faces are often indiscernible. For identification, especially by human, it is useful to render a high-resolution face image from the low-resolution one. This technique is called face hallucination or face super-resolution. In this tutorial we will introduce a simple and effective hallucination technique using eigentransformation.
Finally, we will show a number of demos on face recognition, sketch generation and recognition. The outline and timetable of the tutorial is as follows.
2. Review of Traditional Recognition Algorithms
3. A Unified Framework for Face Recognition
3.1 Unified framework for subspace analysis
3.2 Unified subspace face recognition
4. Face Sketch Recognition
4.1 Sketch transformation
4.2 Sketch recognition
4.3 Face hallucination
Xiaoou Tang received the B.S. degree in 1990 from the University of Science and Technology of China, Hefei, China, and the M.S. degree in 1991 from the University of Rochester. He received the Ph.D. degree in 1996 from the Massachusetts Institute of Technology. He is currently an associate professor and director of Multimedia Lab in the Department of Information Engineering of the Chinese University of Hong Kong. His research interests focus on video processing and pattern recognition. Dr. Tang is a Senior Member of IEEE. He is the guest editor of the Special Issue on Underwater Image and Video Processing for IEEE Journal of Oceanic Engineering and the Special Issue on Image- and Video-based Biometrics for IEEE Transactions on Circuits and Systems for Video Technology. He has been invited to give tutorial lectures on face recognition at the IEEE International Conference on Image Processing, 2004, and the Summer School for Advanced Studies on Biometrics Authentication and Recognition, Italy, 2003.