On sequential watermark detection

This paper introduces a new watermark detection paradigm based on sequential hypothesis testing. It is shown that this framework leads to significantly faster watermark detection compared with more commonly used detectors. Fundamental issues such as joint design of watermark encoding and detection are addressed within this framework. Theoretical results are validated through experiments. On an average, the proposed framework can detect watermarks 70% faster than popular detectors. We believe that this could potentially have significant impact on many applications such as video watermark detection and searching for watermarks in a large database of digital contents.