Exchanging faces of humans in video is a challenging research problem which requires successful combination of various methods for face detection, model creation, model parameter extraction and realistic face rendering. Although each of these methods are studied individually by many researchers from various disciplines of computer science, the whole problem, exchanging faces in video photorealistically, is still open for further research due to the complexity involved in creating a fully automatic system. Most of the existing methods that deal with similar problems make use of either markers put on the face of the subject or more than one view of the subject from different angles. Methods that are capable of creating automatic face models are computationally complex and are not suitable to be used in real-time applications. In this thesis, we plan to create a framework for photo-realistic and fully automatic exchange of faces of human subjects in video with an extension to real-time processing for augmented reality applications.

DATE: 20 March, 2007, Tuesday@ 13:40
PLACE: EA 409