Digital watermarking is the concept of permanent and imperceptible tagging of video data. User-specific watermarking embeds tracking information that accurately identifies the last legal recipient of a video. This type of watermark does not enforce content use restrictions but it enables identification of sources of content abuse, acting as a deterrent and encouraging responsible consumer behavior. These types of techniques represent a way to maintain control over video content while permitting convenient content consumption that can compete with free but inconvenient and illegal content distribution.
Required is a secure, robust and imperceptible marking technology that does not degrade consumer experience and yet establishes a reliable trace that survives attacks or passages through the analog hole. This paper presents our implementation approach, optimized to create a commercially significant capability incorporating new concepts tailored to digital video.
A key element of the approach may appear to be counter intuitive, in that the embedded information is not designed to be machine readable but it can be extracted into a human readable form. While it requires human interaction, it allows use of well honed human perception processing to aid the extraction task - which is simply superior to machine recognition and thereby solves the problem of content synchronization. It overcomes the challenges of content misalignment that fools machine readable approaches, which are often unable to read the mark after small geometric content transformations.
This paper outlines the research results that enable human readout and the challenges to make the mark robust yet invisible. The result allows for wide content distribution, ensures the mark is secure and can establish strong evidence linking the content to a consumer while maintaining privacy.