Constant-rate eye tracking and animation for model-based-coded video

The eyes comprise an important expressive facial region, making their animation essential for successful communication. The authors report on the performance of a new algorithm for eye tracking and animation based on the empirical observation that the eye blinks at a roughly constant rate. Experiments show that the algorithm achieves a 100% success in emulating the blinks occurring at the encoder. The algorithm conforms with the MPEG-4 SNHC approach in allocating description to objects based on their level of detail within the scene, as well as allowing the eyes to be animated using a downloaded script that is attached to them. The approach is appealing for very low bit-rate applications as it achieves a potential bit-rate saving with good subjective quality.