Host-Cooperative Metadata Embedding Framework

This paper presents a new metadata embedding framework for multimedia content where the metadata and its related information are registered in a database. A watermarking method based on the framework computes the metadata that provides content with the smallest perceptual artifact by way of extracting the metadata from the original unwatermarked content. This can reduce the watermark power at the embedding because the content itself naturally contains the metadata. In the case that the extracted metadata has already been registered in the database, several bits in the metadata are inverted so as to avoid the double metadata registration and also detect it with probability higher than the other unregistered metadata.