Global, simultaneous registration of multiple data sets is common problem in medical image processing. With the rising of 3D methodologies and of 3D functional imaging, it is likely that the importance of this issue will increase in the near future. Despite the incredible number of research paper published in the image registration field, a low attention was pointed on the problem of simultaneous registration of multiple images. The common used approach is heuristic and based on the reduction of the problem to a series of registration of image pairs. This practical approach is fair to be optimal and a robust theory of global image registration is not available today. When more than two data sets had to be registered each other simultaneously, the complexity of the registration problem increase and new registration metrics may be involved. In theory, is possible to extend the similarity metric currently used in image registration, as the mutual information, to more than two images. However, registration of multiple data sets is usually based on a series of image pair registrations in order to reduce computational complexity and to exploit available registration algorithms. In this paper, we define the general global registration problem and explore different approaches to solve it. Finally, we introduce the evolutionary computation approach that is promising in address the optimization constraints introduced by the global registration problem.
Simultaneous registration of multiple images: applications to multimodal and dynamic medical imaging
Positano V;Casciaro S;Landini L
2007
Abstract
Global, simultaneous registration of multiple data sets is common problem in medical image processing. With the rising of 3D methodologies and of 3D functional imaging, it is likely that the importance of this issue will increase in the near future. Despite the incredible number of research paper published in the image registration field, a low attention was pointed on the problem of simultaneous registration of multiple images. The common used approach is heuristic and based on the reduction of the problem to a series of registration of image pairs. This practical approach is fair to be optimal and a robust theory of global image registration is not available today. When more than two data sets had to be registered each other simultaneously, the complexity of the registration problem increase and new registration metrics may be involved. In theory, is possible to extend the similarity metric currently used in image registration, as the mutual information, to more than two images. However, registration of multiple data sets is usually based on a series of image pair registrations in order to reduce computational complexity and to exploit available registration algorithms. In this paper, we define the general global registration problem and explore different approaches to solve it. Finally, we introduce the evolutionary computation approach that is promising in address the optimization constraints introduced by the global registration problem.File | Dimensione | Formato | |
---|---|---|---|
prod_135254-doc_99285.pdf
solo utenti autorizzati
Descrizione: Capitolo pubblicato
Dimensione
379.67 kB
Formato
Adobe PDF
|
379.67 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.