Let us now consider two images of the same modality and of the same patient, but in a different position. We extract extremal lines on both images. The problem is to put into correspondence the two sets of lines (the model and the scene), which is often called the matching step, and to compute the best rigid transformation that superimposes the matched lines.
It is important to note that a global registration algorithm, for instance superimposing the barycenters of all points and the inertia axes, will often fail because of the occlusion. Indeed, the images being taken in different positions, the region of interest are frequently different in the two images, leading to crest lines and extremal points present in one image and not in the other. The image noise will also induce the extraction of spurious lines and points in different parts of the two images.
Was this article helpful?