Image Processing Reference
In-Depth Information
Tolerance nearness measure (tNM), introduced in (Henry, 2009) is based on the idea that
if two images are similar, tolerance classes in the union set of those images will have similar
number of subimages from each image. tNM between two images (Henry, 2009; Henry and
Peters, 2009a) is based on definition 9.4.2 and is defined as follows:
Suppose X and Y be sets of perceptual object(for example two images), and Z = X∪Y be
the union of X and Y . Let [z /
= B,ε ] ⊆X
and [z /
= B,ε ] ⊆Y
denote the portion of the elementary
set z /
that belongs to X and belong to Y respectively. Then,
= B,ε
[z /
= B,ε ] ⊆X ,{x∈z /
= B,ε |x∈X}⊆z /
= B,ε ,
(9.11)
[z /
= B,ε ] ⊆Y ,{y∈z /
= B,ε |y∈Y}⊆z /
= B,ε ,
(9.12)
z /
= [z /
= B,ε ] ⊆X ∩[z /
= B,ε ] ⊆Y .
(9.13)
= B,ε
Tolerance nearness measure (tNM) is defined as the weighted average of the closeness
between the cardinality (size) of set [z /
= B,ε ] ⊆X
and the cardinality of [z /
= B,ε ] ⊆Y
where the
cardinality of z /
is used as the weighting factor.
= B,ε
× X
z / = B,ε
1
min(|[z /
= B,ε ] ⊆X |, |[z /
= B,ε ] ⊆Y |)
tN M =
X
= B,ε ] ⊆Y |) ×|z /
= B,ε |.
(9.14)
max(|[z /
= B,ε ] ⊆X |, |[z /
|z /
= B,ε |
z / = B,ε
9.6
Perceptual Image Analysis System
This section explains some implementation details of the perceptual image processing system
discussed in the previous sections. The system has been implemented using the latest version
of C++ .Net(visual studio 2008), combining both managed and non-managed C++ coding
techniques. The object oriented structure of the program makes it easier to make any
further changes and improvements to the program; and the choice of using C++ language
rather than Matlab makes the program much faster. Figure 9.4 shows a view of the system.
The GUI has different parts that are designed for different purposes. Image panel 1 and
image panel 2 are two of the most important panels, where you can see two images that
will be compared and processed. Image panel 1 shows the first image and image panel 2
includes the second image. The images can be selected by clicking on the 'select image'
buttons. Gray scale of each image will be shown in the second image box of panels.
Third image box in each panel shows the covering (all tolerance classes) for the selected
image. This can be seen after clicking the 'Calculate Tolerance Classes' button by user. In
the case that processing of more than two images is needed and for the sake of simplicity,
user can choose a folder on their computer and give its path to the system using the dialog
box and browse button at the top of 'Image data base' panel. The system will perform the
comparison on each pair of images on the specified folder.
Another advantage of this system, from the user friendly point of view, is the option to
have up to 12 different images loaded to the system at the same time and being able to
choose which ones are the ones being compared and processed at the moment. The image

Search WWH ::

Custom Search