Cryptography Reference
In-Depth Information
7.4 Statistically Adaptive Detection Using Inferential
Statistics
Video providers subject watermarked pictures to various kinds of video
processing. These include compression using MPEG or other compression
technology, resizing, and filtering. These video processing procedures can be
exploited by illegal users to remove the embedded information. The water-
marks should be robust enough to be reliably detected after any of these
processes. Watermark survivability is essential, and methods for improving
WM survivability have been actively studied. Methods using redundant cod-
ing [17, 18, 19] and spread spectrum coding [20, 21] of the embedded informa-
tion have been established. Methods for embedding information in elements
that will be negligibly affected by the expected image processing have also
been developed [22, 23].
Redundant coding is a basic method for improving WM resistance to video
processing without degrading picture quality. It embeds watermarks repeat-
edly in every frame or region. The embedded watermarks are accumulated and
detected during watermark detection. The chance of error is reduced. This
accumulation, may be problematic after video processing that removes WMs
from specific frames or regions. This is because a WM signal could be atten-
uated due to attempted accumulation from regions from which the WMs had
been removed during video processing. The detection method should there-
fore be used to detect WMs only from those frames and regions where WMs
remain.
The statistically adaptive detection technique sets up a scale for estimat-
ing the bit-error rates of the WMs for each frame region by using statistical
properties of motion pictures. It uses those rate to determine the regions in
which to accumulate so as to prevent watermark signal attenuation.
7.4.1 Conventional Methods
Redundant coding methods can be classified into two types.
(a) Redundant Coding within a Frame: Watermark embedding is done by di-
viding the frame picture into several regions and applying the same wa-
termark to each region. In WM detection, the WMs are extracted by
accumulating all the divided pictures in a frame [17, 18].
(b) Redundant Coding over Frames: WM embedding is done by applying the
same watermark in each consecutive frame of the video. In WM detection,
the WMs are extracted by accumulating all the frames or a specific number
of sequential frames [18, 19].
Among these methods, the one proposed by Kalker and coworkers [18] is
a basic WM scheme using both types. We briefly describe this method and its
problems. For ease of understanding, we first introduce the 1-bit-WM scheme
and then describe the multiple-bit-WM scheme.
