Literature Reviews - Segmentation of Hand Bone for Bone Age Assessment - page 26

Thresholding

Manual
Threshold
selection

Automated
Threshold
selection

Single
Threshold

Multiple
Thresholds

Computational
Efficiency

Global
Thresholding

Local
Thresholding

Dynamic
Thresholding

Adaptability

Fig. 2.5 The taxonomy of Thresholding segmentation

Various Thresholding segmentation techniques have been reviewed in this
sub-section. Each category of thresholding with different schemes of threshold
selection and number of thresholds has different capability in segmentation task.
Generally, Global thresholding has the highest computational efficiency followed
by local thresholding and dynamic thresholding. Global thresholding, however, has
the lowest adaptability to deal with uneven illumination, followed by local thresh-
olding and dynamic thresholding. The experiments on hand bone segmentation
using aforementioned thresholding methods showed that thresholding methods can
hardly produce promising resultant segmented hand bone; Taking no consideration
of object's features in thresholding technique is probably the reason. Therefore, in
next sub-section, the edge-based segmentation methods would be explored. Lastly,
the taxonomy of Thresholding segmentation is summarized using the Fig. 2.5 .

2.4 Edge-Based

Thresholding segmentation discussed in previous sub-section postulates that fore-
ground and background can be distinguished by certain range of pixels intensity.
edge-based segmentation, differently, postulates that foreground and background
are distinguished by a closed boundary formed by meaningful edges. Edges are