LINEBURG


<< . .

 23
( 28)



. . >>

46


53


61


69


77


84


92


10


11


11


13


24




Figure 3. Histogram of the slice 104

According to the analysis of the above histograms, the whole CT images can be
divided into 3 regions as the following. The gray level of the first region is between 0 and
360, where the main meterials are air and the mattress of CT bed with minimum density. The
second region is mainly composed of organic soil, plastic container and light impurity with
smaller density and maximum volume, of which the gray level is between 360 and 800.The
root system and weighty impurity belong to the third region with the gray level above 800. It
was therefore possible to preliminarily determine that the threshold value of the root system
is over 800. Figure 4 shows the raw image of the silce 104 and its initial segmented result.


186
ISBN 5-88890-034-6. 1.




(a) (b)
Figure 4. CT image of the slice 104 and its primary segmentation
(a) raw image of the slice 104 (b) primary segmentation of root system (T>800)




In addition to the root system, the primary segmented regions also include a great deal
of weighty impurity and other components. So it is necessary to do more precise
segmentation.
In order to get a precise segmentation, the typical slices with clear root regions were
selected from the slice 160 to the slice170, and the histograms of the typical root regions and
the rhizospheres including root system were drawn(Figure 5 and figure 6). At the same time,
the whole histogram and the histogram of typical slices above with the gray level of more
than 800 were further analyzed. Figure 7 is the histogram (T> 800) of the slice 165.
12



10


8



6



4


2



0
15 7

25 7

30 6

34 1

37 7

41 2

44 5

47 6

50 8

53 9

57 3

63 0

71 8

82 0

92 7

100 0

104 8

108 4

111 8

115 0

118 0

121 2




Figure 5. Histogram of the rhizosphere including root system in the slice 165




187
ISBN 5-88890-033-8. . , 2005.


9

8


7

6

5

4

3

2

1

0




436

604

702

755

828

890

938

975

1011

1036

1056

1077

1095

1114

1131

1148

1164

1180

1196

1214

1231
Figure 6. Histogram of a root region in the slice 165

30



25



20



15



10



5



0
800

829


858

887

918

949


981

1012

1042

1072

1101


1130

1159

1188

1218


1273

1585

2000

2533

2899
Figure 7. Histogram of the slice 165

It is clear that there are two peaks in figure 5, and the root system is just in the right
peak of which the gray level is more than 800. In fact, the values of the pixels in the right
peak are between 930 and 1240. Figure 6 is the histogram of a root region in the slice 165,
where the pixel values in the root region are between 485 and 1245. By contrasting figure 5
with figure 4, it is obvious that there is no clear boundary between the gray levels of root
system and its media, of which the distributing range are intercrossed each other. According
to the result from statistic analysis, the pixels of root system whose gray level is in the range
from 930 to 1230 are 95% of all pixels in the root region. And combining with the whole
histogram above and the histogram of the typical slices, it can be seen that the range of
threshold value for the precise segmentation of root is from 930 to 1230. The effect of
segmenting CT images with the threshold 930˜1230 was shown in figure 8.




188
ISBN 5-88890-034-6. 1.




(a) (b)
Figure 8. Effect of segmenting with the threshold 930˜1230:
(a) Segmentation of the slice 104 (b) 3D display of segmenting effect

2.2 Mathematical Morphological Processing
It was shown in figure 8 that there was still some impurity after 3D threshold segment-
ing. Most of the impurity was some spots whose density was very close to root system, some
others may be caused by the local reaction of CT imaging. In order to improve the precision
of the subsequent analysis and measurement, it is necessary to make further segmenting for
the images. Therefore, mathematical morphological processing was used for the images seg-
menting. Mathematical morphology is a set-theoretical algebra consisting of two fundamental
operators, dilation and erosion. A binary signal can be considered a set X, and erosion and di-
lation then correspond to Minkowski addition and subtraction with respect to another set E
called the structuring element. Here we use the notation
X E={?+?: ? X, ? E}
for dilation of a set X by structuring element E. Erosion is then the dual operator of di-
lation
X?E=(Xc E )c
Where Xc denotes the complement of X, and E denotes structuring element E
reflected about the origin.Further morphological operators are formed as combinations of di-
lation and erosion. The open operator is defined by
X0E = (X?E) E
And its dual, the close operator by
X?E = (X E)?E
Open operating can erase the region less than structuring element and cut off the grac-
ile connective band between larger regions. The operating was selected as the main process of
mathematical morphology for the CT images after 3D threshold segmenting and binarization.
The effect of the operating is shown in Figure 9.




189
ISBN 5-88890-033-8. . , 2005.




(a) (b)
Figure 9. Effect of mathematical morphological processing for the slice 104
(a) before processing (b) after processing

2.3 Segmenting based on the geometrical feature of root system
In order to remove most of the impurity whose density is very close to root system, a
segmenting algorithm based on the geometrical feature was developed. Special algorithm was
designed by the following principles: the spatial location of root system between adjacent
slices and the change of its sectional shape and area are both continuous. As the root regions
in the CT images of rhizome are very clear and easy to be recognized, so the first step of the
algorithm is to setup the basement of recurrence with the analysis of the rhizome CT images.
Then according to the spatial continuity of the root system, it is possible to find out the re-
gions of root system downwards slice by slice and weed out the pixels of the impurity. Figure
10 shows the effect of geometrical segmentation.




a (b)
Figure 10 Effect of geometrical segmentation
(a) Slice 104 after segmentation (b 3D display of segmenting effect

It is obvious that almost all impurity was removed after geometrical segmentation.
Compared with the medical images segmenting methods available, the interactive segmenting
algorithm get better segmenting effect.



190
ISBN 5-88890-034-6. 1.



3 CONCLUSIONS
On the basis of analyzing the gray-level histograms of the whole images and typical
slice images, the distribution of gray-level for the whole CT images was distinguished with
three different regions. There is no clear boundary between the gray levels of root system and
its media, of which the distributing range are intercrossed each other. According to the result
from statistic analysis, it was determined that the range of threshold value is from 930 to 1230
for the 3D threshold segmenting of the root.
After further morphological processing and geometrical feature-based segmenting,
almost all the impure pixels with its density close to root system were removed from the CT
images of the root. The result of programming experiment showed that the integrated algo-
rithms is an more effective method for segmenting the images.

Acknowledgements
This research was supported by the National Natural Science Foundation of China un-
der Grant
No. 60375005.

REFERENCES
Cui Yi. 2000. Image processing and analysis—mathematical morphological methods
and its applications. Beijing: Beijing Science Press.
He Bin, Ma Tianyu, et al. 2002. Visual C+ + digital image processing (2nd edition).

<< . .

 23
( 28)



. . >>

Copyright Design by: Sunlight webdesign