Literature Reviews - Segmentation of Hand Bone for Bone Age Assessment - page 17

2.3 Thresholding

Thresholding is one of the earliest image segmentation techniques, and yet it
remains to be the most widely applied segmentation technique attributable to its
simplicity and intuitiveness [ 32 ]. Thresholding segmentation is normally con-
ducted in spatial domain based on the postulation that both object and background
are represented by different range of pixel intensity [ 33 ]. Basically, there are three
categories of thresholding: Global thresholding, local thresholding and dynamic
thresholding [ 34 ].

2.3.1 Global Thresholding

Undoubtedly, the simplest method in thresholding techniques to segment an image
is through Single Global thresholding: this technique based on the concept that if
object in the image and other object or background are mutually exclusive in terms
of intensity range, then it could be separated in different partition using a single or
multiple values of pixels intensity [ 35 ]. In the case of single threshold, it can be
represented as follows:

g 1
g 2

IF
IF

f ( x , y ) < T
f ( x , y ) ≥ T

f ( x , y ) = (2.1)

where g if denotes the group of pixels that represents an object or background; if a
pixel value is less than T, which is the threshold value, then it is grouped into g 1 ; if
a pixel value is greater than T, then it is grouped into g 2 . The f ( x , y ) denotes the
image pixel intensity in 2D gray-scale image in coordination ( x , y ) . The concern
of the technique is to classify an image into object and background; this type of
grouping is called binarization.
The single thresoholding depends on the T. This T value determines the inten-
sity range of an object and the intensity range of the image background. For
instance (if the object is brighter than the background), if an pixel intensity value
is greater than the threshold value, then the pixel will be classified as object; for
the pixels which possesses intensity value less than or equal the threshold value,
they will be considered as background. This kind of thresholding method is con-
sidered as 'threshold above'; another type is 'threshold inside' where the object
value is in between two threshold values; similarly, another variants is 'threshold
outside' where the values in between the two threshold values are classified as
background [ 36 ].
The accuracy of thresholding technique in segmentation mainly depends on two
factors: first factor is the property of the image intensity distribution of both object
and background. Thresholding technique performs most efficiently when the
intensity of input image has distinct bi-modal distribution without any overlapping