Cryptography Reference
In-Depth Information
7.2 Video Watermarking
7.2.1 Typical Use
As shown in Fig. 7.1, watermarks (WMs) representing copyright information
are typically embedded into video pictures that are then compressed using
an MPEG encoder. The video provider broadcasts the encoded pictures or
distributes them over a network or on recording media, such as DVD-ROMs.
When the encoded watermarked pictures are received, they are decoded, and
a user can watch them on a TV set or a personal computer. If a user illegally
copies and redistributes the pictures (for example, by placing them on a web
page), the auditor detects the copyright information embedded in the pictures
and notifies the video provider of the illegal copying. The video provider could
even identify the illegal copier if the pictures contained embedded information
identifying the user to whom they had originally been provided.
7.2.2 Overview of Two Techniques
As mentioned above, watermarks must survive video processing and not de-
grade picture quality. That is, they must be detectable even after the pictures
have gone through image-processing procedures such as MPEG compression
and resizing, and they must not interfere with picture enjoyment. These re-
quirements conflict because survivability is ensured by embedding more WMs,
which in turn degrades picture quality. As mentioned above, resolving this
tradeoff has been a major objective of the research on watermarking, and
numerous studies have focused on both watermark embedding and detection.
However, the methods reported so far are not always effective because they
simply utilize watermarking methods developed for still pictures and neglect
the properties of motion pictures. We have been studying video watermarking
in the pixel domain and previously proposed improved embedding and de-
tection techniques using motion picture properties [3, 4, 5, 6]. These motion-
adaptive embedding and statistically adaptive detection techniques resolve
the tradeoff in requirements.
Motion-adaptive embedding [3, 4]: Taking into consideration the properties
of the pictures enables adaptive allocation of the watermarks. Picture
quality is degraded less by embedding the watermarks in areas where they
are easily perceived; survivability is improved by embedding them more
heavily in areas where they are hard to perceive. Conventional methods
consider only the properties of each frame and neglect the inter-frame
properties. We have found that watermarks are less perceptible when em-
bedded in areas where the picture contents are moving, and we have de-
veloped watermark imperceptibility criteria based on motion information.
We also developed an embedding technique that uses this criteria to allo-
cate watermarks to picture areas adaptively.
