therefore worth further study. We believe that informed watermarking offers significant near-term improvements. While proposed codes for informed embedding are computational ef- ficient they are not robust to valumetric scaling. Chen [6] briefly proposed a solution, but further investigation is needed to realize computationally efficient and robust codes.

Handling geometric/temporal distortions in a blind detector remains a dif- ficult problem. However, a number of different approaches have been investigated and incremental progress is being made. A breakthrough is probably not imminent, but progress in this area would lead to significantly more robust systems.

While many papers have illustrated the use of a variety of fidelity models, there has been very little work [15] on how to optimally embed a watermark with fidelity and robustness constraints. We expect this to become a fruitful new area of research.