Image Processing Reference
Figure 9.7. Information expressed in the image space. White and black correspond to
minimum and maximum membership values, respectively. Top left: information on the location
and the approximate shape provided by the atlas. Top right: binary localization constraint
expressing the fact that the caudate nucleus is located in the brain (search area, in black)
and outside the lateral ventricles in white. Bottom left: prior radiometric knowledge.
Bottom right: relative directional relation “to the left of the lateral ventricles”
because of the anatomical variability among individuals. Finally, fuzzy fusion which
leads to recognition can take advantage of the many operators which make it possible
to adequately model the relations we know between these elements of information.
Modeling is achieved by representing, in the image space, each element of infor-
mation by a fuzzy set. Fuzzy areas of interest are thus defined for each type of knowl-
edge and their fusion allows us to focus our research in a increasingly limited area. We
will illustrate this process with the recognition phase of the caudate nucleus. At this
phase of the recognition procedure, three objects have been segmented: the brain and
the two lateral ventricles. Figure 9.7 illustrates the representation of the knowledge