Abstract
The requirement for improved picture quality in videophone and videoconference systems operating at low bit-rates has stimulated interest in model-based image coding. In this paper, two model-based coding techniques are described which are capable of producing either improved picture quality at bit-rates around 64 kbit/s or acceptable picture quality at bit-rates far lower than 64 kbit/s. The first technique produces facial expressions by using feature code-books; the second technique produces facial expressions by distorting an underlying three-dimensional model. The problems of image analysis and synthesis, which are concomitant in model-based coding, are discussed.

This publication has 1 reference indexed in Scilit: