Image Registration Methods in MRI - Advance Image Processing in Magnetic Resonance Imaging - page 187

Image to Be
Registered

ROTO-TRANSLATION

INTERPOLATION

Reference
Image

REGISTRATION
METRIC
EVALUATION

T

TEST

Registered
Image

OPTIMIZATION

FIGURE 7.1 Flowchart of the general registration problem between two images.

a local (not global) optimum. However, this problem can be overcome by selecting
an initial orientation close to the correct registration, as we have seen in the
previous paragraph. We will assume in the following text that the global optimum
is obtained at the correct registration transformation. Another important quality
of the similarity metric is the computational complexity that affects the time
required to perform the registration.
Generally, three types of similarity metrics have been proposed in image
registration [10]. They are based on corresponding points, corresponding surfaces,
and corresponding image intensities. We can group the first two methods in one
and summarize the two main approaches to registration as follows:

1. Similarity measure by extraction of some geometrical features from

the two images: The extracted features are compared, and a similarity