The Design And Implementation Of Registration Algorithm In Ct/mri Image

Medical image registration is an important and rudimental study subject in the area of medical image processing. It is the basic step for multi-image analysis of the same scenc, and it is also the foundation of medical image fusion, restoration and construction of standard altas, etc. So, registration is an important study subject in medical image processing. Because of the complexity of human body structure, we can used rigid registration techncology to registed the tissues that can be seen as rigid object, but many tissues are non-rigidity. Simple rigid registration cannot show the local details of body complex structure. But these details are the important gist of clinical diagnosis. So it needs non-rigid matching technology with higher freedom.This paper takes multual model medical image as data sources, and researches and develops non-rigid registration algorithms according to practical needs of clinical diagnosis, design and implementation a rigid registration algorithm based on corner Features and a non-rigid registration algorithm based on multual information controlled free form deformation, obtaining good experimental results.First, some pretreatments such as:gray conditioning, smoothing, coordinate mapping, interpolation processing, ect are taken to remove the influence of the defferences between the two images intensity, nosie, sizes, etc. In order to registration medical images that contains rigid tissues, a rigid registration algorithm based on corner Features is designed, which is based on seriously study of principal component analysis and harris corner detection, implemented automatic rigid registration between rigid medical images. In order to registration medical images that contains non-rigid tissues, a non-rigid registration algorithm based on multual information controlled free form deformation is designed, which is based on seriously study of free form deformation, multual information and B-spline technic, implemented efficient automatic non-rigid registration between non-rigid medical images.Extensive test data show that the algorithms in this paper realize accurate and efficient automatic registration of medical image and is simple and convenient in operation. It can be very good to meet the actual needs.