Biomedical Engineering Reference
In-Depth Information
100
Number of pixel vs Pixel intensity Bin
Multiple Gaussian Model
90
80
Third Inflation point
70
60
50
40
Second Inflation point
30
20
10
0
0
50
100
150
200
250
8-bit Gray-scale Pixel Intensity Bin
Fig. 3.22 The histogram of segmented bone image of Fig. 3.20 a and its interpolated Gaussian
graph with second inflation point and third inflation point
For instance, the inflection points of the interpolated Gaussian graph of the arbi-
trarily selected segmented bone image of Fig. 3.20 a, are 144 and 182. Thus, those
pixels belong to this newly defined group, which are pixels with pixel intensity
between 144 and 182, are going to be analyzed, whether it is to be eliminated or
remained. Pixel intensity value more than 182 is considered as compact bone group
and pixel intensity value more than 144 is considered as background group. The his-
togram, interpolated Gaussian model and the inflection points are shown in Fig. 3.22 .
For segmented bone image of Fig. 3.20 b, the inflection points of the interpo-
lated Gaussian graph are 197 and 242. Thus, those pixels belong to this newly
defined group, which are pixels with pixel intensity between 197 and 242, are
going to be analyzed, whether to be eliminated or restored. The histogram, inter-
polated Gaussian model and the inflection points are shown in Fig. 3.23 .
3.5.2 Hand Bone Edge Detection Technique Using Entropy
Owing to the nature of hand bone image, conventional image processing tech-
niques are unable to detect the edges easily (this statement will be justified in next
chapter). Therefore, we propose a technique where the by-product of anisotropic
Search WWH ::




Custom Search