Biomedical Engineering Reference
In-Depth Information
Floating Image
Initial Transform
Transform
Update Transform
Reference Image
Transformed Floating
No
Compute
Mutual Info
Mutual Info
Maximized?
Yes
Final Transform
Figure 4.1:
Flow-chart of image registration by mutual information maximi-
zation.
Our implementation is based on Java, using a lot of Java 2D graphics and
image processing capabilities. Java 2D API includes a set of classes that can be
used to create high quality graphics. It has features like geometric transforma-
tion, alpha composition, and image processing (see [33, 34]).
4.2.2
Transformation
In image registration different transformations can be considered: rigid-body
transformation, affine, and nonlinear warping, with increasingly technical dif-
ficulty. In this chapter we consider affine transformation with a scaling factor,
rotation angle, and two translational offsets. It is reported that the affine trans-
form is sufficient for most retinal registration [23].
