The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
Figure 4. The flow of detecting watermark 
Where lfl{k) represents the original watermark, 
rn\k) represents the detected watermark, VI represents the 
length of bit for watermark. 
Here the original data of coordinate points are needed to record 
so as to detect the watermark. 
5 THE EXPERIMENTS 
We now present the experimental results for the proposed 
watermarking algorithm. The watermark is shown in Figure 1, 
and the original data to embed watermark is the contour data 
including 160182 coordinate points with the scale 1:250000. 
5.1 Visibility 
Figure 5 shows the embedded watermark data overlapped with 
the original data, where the solid lines is the original contour 
data and the hidden lines is the embedded watermark data. From 
the comparison of the two kinds of data in Figure 5, it can be 
known that the proposed watermarking algorithm is with good 
imperceptibility. 
Figure 5 the imperceptibility for the embedded watermark data 
5.2 The precision analysis 
There are 160182 coordinate points in the original and 
embedded watermark data. We compare the absolute error 
between two kinds of data. The comparison results list is in 
Table 1. 
Absolute error C 
C=0 
C=1 
C=2 
02 
Number of points 
155024 
4204 
954 
0 
Percent 
96.78% 
2.62% 
0.60% 
0 
Table 1 The error between original and embedded watermark 
data 
From Table 1, it can be known that there is no error for the 
96.78% of total data. For the points with the error, the error is 
allowable (<=2), and the quality of data with watermark can 
satisfy the actual applications. Hence the proposed 
watermarking algorithm is with high precision. 
5.3 The robustness 
In the subsection, the robustness for the embedded watermark 
data is studies, and the watermark with intensity 1 and max 
error 2 is detected after meeting the different attack. 
(1) No change 
If there is no any change for the embedded watermark data, the 
watermark can be detected wholly. 
(2) Compressing 
If compressing the embedded watermark data, the watermark 
also can be detected and the result is shown in Figure 6, where 
the correlative coefficient is 0.86689. 
Figure 6. The detected watermark by compressing 
(3) Deleting points 
We delete some points randomly from the embedded watermark 
data, and then detect the watermark from the data. The result is 
shown in Figure 7, where the percent of deletion is 10% and the 
correlative coefficient is 0.904580. 
Figure 7. The detected watermark by deleting points 
(4) Noise attacking 
If attacking the embedded watermark data by uniform noise, the 
watermark also can be detected and the result is shown in 
Figure 8, where the correlative coefficient is 0.857824.