The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
4. RESULTS
Many experiments are carried out using several images oriented
to different objectives. The results show that this method is very
useful when extracting edges between objectives and the
background. Following are four of them to show how this
method works.
Images with initial curve are on the upper side and the results
are on the underside. The first experiment (Figure 2) is to
extract the edge of a lake. From the image we can see that the
edge is very clear and the water is distinct from the land, if we
use threshold segmentation, the rivers will influence the result,
we will do some extra work to remove edges we do not need.
By using level set method, we can see from the result that it
successfully extracts out the edge without facing those problems.
Figure 3 shows the second experiment, this experiment is to
extract waterline of tidal flat, Because of the influence of the
tidal, the shape of the waterline is very complicated, and this
make the curve evolves a longer time than the first experiment.
When this method extracts the waterline, we also can see some
Figure 2. Extract lake edge. Upper: image with initial curve.
Underside: result curve
spots and they also exist in Figure 4 and Figure 5. The reasons
of producing these is partly because the C-V model which we
used. But it is not a big problem, we can just track the line we
want and do not care about them. In Figure 4, we assume a
situation that there are several homogenous areas in an image,
but we only want to extract one of them, the image was cut from
an image of a city, there are several green areas in this in image,
then we want to extract one of them and we can see that this
method works very effectively. The fourth experiment is to
extract out waterline of Jiuduansha Shoal in Changjiang Estuary
(Figure 5), Jiuduansha Shoal located in the Turbidity Maximum
zone in Changjiang Estuary, and high suspended sediment
concentration greatly enhances the reflection of the water body
in this area. So in some area, the DN value from water to
mudflat changes gradually which makes it hard to determine the
exact location of the waterline, but the idea of minimize energy
of C-V model can solve this problem in an reasonable way.
From this experiment, we can see another advantage of this
method which is that when the curve evolves, it will split when
it meets several targets.
Figure 3. Extract waterline. Upper: image with initial curve.
Underside: result curve