11 Information Hiding for Spatial and Geographical Data
243
distortions that could be perceived in terms of quality degradation of the resulting
watermarked digital map against the original one. Similarly, the vector maps show
this feature; in fact, maps are represented in terms of geometric features and embedding a watermark may change the geometry of the map.
In [18], this issue is analyzed and a blind watermarking algorithm is provided:
the watermark is imperceptible, so the resulting map is of high quality. The basic
idea is to add local small noise signals at the boundaries between the homogeneous
regions in order to obtain the watermarked map. Also, this watermarking algorithm
is robust against cropping and shifting attacks since the watermark is added locally.
In the following, we illustrate how this watermarking algorithm works. The region boundaries are extracted by means of a segmentation algorithm. The watermark
is chosen from the set {−1, 0, 1} to avoid possible boundary points shifting in the resulting watermarked map. Indicating with L w , the watermark length, the following
formula states that both the endpoints of the original and the modified boundaries
will coincide.
L w −1
n=0
w n = 0
(11.3)
In general, it is possible to constrain the endpoints of both the original and the modified boundary to coincide at regular intervals, as follows:
n+l−1
n=m
w n = 0
(11.4)
with m = 0, l, 2l, . . . , L w . The watermark detection is performed by means of correlation between the watermark and the watermarked map. The watermark is assumed
to be present if the correlation value is over a specified threshold.
According to the experimental results, a watermark of 75 elements is added to a
color-mapped chart image of resolution 10000 × 3000, which represents
hydrographic resources. Even if the watermark is embedded unobtrusively, the watermarking algorithm is shown to be robust against main attacks, including the JPEG
compression and the addition of Gaussian noise. Regarding the JPEG compression
attack, the watermark can be detected when the quality factor is over 20; this result is acceptable since digital maps should preserve their quality, otherwise they are
useless.
The watermark embedding comes at the price of little distortions in the digital
map. Watermarking developers try to reduce these distortions to make the watermark
imperceptible. Another requirement of watermarking is the usability of watermarked
content. This requirement is very important in those applications where the watermarked content serves as input for further operations. In GIS applications, watermarked maps may be visualized or processed by common GIS operations (selection,
buffering, overlay, vector model conversion, classification) to extract the desired geographical and spatial information.
In [13], the authors take into consideration the impact of watermarking on digital
maps for further processing, especially for classification of satellite images. GIS allow users to integrate different sources of geographical and spatial data. The remotely
243
distortions that could be perceived in terms of quality degradation of the resulting
watermarked digital map against the original one. Similarly, the vector maps show
this feature; in fact, maps are represented in terms of geometric features and embedding a watermark may change the geometry of the map.
In [18], this issue is analyzed and a blind watermarking algorithm is provided:
the watermark is imperceptible, so the resulting map is of high quality. The basic
idea is to add local small noise signals at the boundaries between the homogeneous
regions in order to obtain the watermarked map. Also, this watermarking algorithm
is robust against cropping and shifting attacks since the watermark is added locally.
In the following, we illustrate how this watermarking algorithm works. The region boundaries are extracted by means of a segmentation algorithm. The watermark
is chosen from the set {−1, 0, 1} to avoid possible boundary points shifting in the resulting watermarked map. Indicating with L w , the watermark length, the following
formula states that both the endpoints of the original and the modified boundaries
will coincide.
L w −1
n=0
w n = 0
(11.3)
In general, it is possible to constrain the endpoints of both the original and the modified boundary to coincide at regular intervals, as follows:
n+l−1
n=m
w n = 0
(11.4)
with m = 0, l, 2l, . . . , L w . The watermark detection is performed by means of correlation between the watermark and the watermarked map. The watermark is assumed
to be present if the correlation value is over a specified threshold.
According to the experimental results, a watermark of 75 elements is added to a
color-mapped chart image of resolution 10000 × 3000, which represents
hydrographic resources. Even if the watermark is embedded unobtrusively, the watermarking algorithm is shown to be robust against main attacks, including the JPEG
compression and the addition of Gaussian noise. Regarding the JPEG compression
attack, the watermark can be detected when the quality factor is over 20; this result is acceptable since digital maps should preserve their quality, otherwise they are
useless.
The watermark embedding comes at the price of little distortions in the digital
map. Watermarking developers try to reduce these distortions to make the watermark
imperceptible. Another requirement of watermarking is the usability of watermarked
content. This requirement is very important in those applications where the watermarked content serves as input for further operations. In GIS applications, watermarked maps may be visualized or processed by common GIS operations (selection,
buffering, overlay, vector model conversion, classification) to extract the desired geographical and spatial information.
In [13], the authors take into consideration the impact of watermarking on digital
maps for further processing, especially for classification of satellite images. GIS allow users to integrate different sources of geographical and spatial data. The remotely