Database Reference
In-Depth Information
Fig. 2.
General Digital Watermarking Process
concepts are used to watermark the data in digital watermarking (compared to database
watermarking). We depict the general watermarking process in Figure 2. In general, a
digital watermark system consists of two components: First, the
embedder
,whichis
responsible for embedding the watermark in the multimedia data. Second, the
detector
,
which determines whether a watermark is present or not [9]. In the case that an invert-
ible watermark is used, the system is extended by a
watermark extractor
that restores
the original data. For embedding a watermark, the embedder needs three inputs: The
original, unwatermarked work (i.e., the digital data) the information (data) to embed
and a key. As a result, we create the watermarked data that serves as input for the de-
tector. As with database watermarking, a digital watermark system can have different
characteristics such as visibility, robustness, fragility, transparency, capacity and invert-
ibility. The characteristics of a specific watermark system depends on the application
the watermark is used for and can be chosen according to the system requirements.
Amongst others, proof of ownership and content authentication are common applica-
tions of watermarking. For a detailed overview of watermark application areas we refer
to Dittmann et al. [12].
2.3
Invertible Watermarking Techniques
A major problem of watermarking techniques is that watermarked data always con-
tains a particular amount of noise (i.e., the watermark) depending on the watermarking
scheme [14] . Consequently, watermarked data used in complex transformations (e.g.,
image quality enhancement) may lead to incorrect or at least biased results. To over-
come theses limitations,
invertible
watermark approaches that can extract the watermark
and restore the original data have been recently proposed in the literature. According to
Feng et al., there are three major approaches to embed invertible watermarks: (1) Com-
pression based approaches, (2) Difference expansion, and (3) Histogram shifting [14].
Compression based approaches.
These approaches compress some parts of the image
and use the gained space to embed the watermark message.
Difference expansion.
These techniques either use pixel pairs or blocks of pixels. The
difference of colour values among the used pixels is expanded to create space for
the embedding procedure. This procedure is invertible. Subsequently, we will ex-
plain one of these techniques in more detail.
Histogram shifting.
A histogram shifting scheme first segments an image into several
blocks. Then, it computes the histogram of pixel colour values for each block. To
embed an a one, theses techniques change the sequence of histogram bins according
to a certain algorithm.
