Algorithm And Implementation Based On Non-rigid Medical Image Registration

Medical image fusion, a research field that involves information science, computer science and medical imaging, is a new hot spot in medical image processing.Image fusion has three important parts, including image pre-processing, forming fusion image by fusion criterion, estimation on quality of image. Image pre-processing includes de-noise, enhance and registration of image. Image registration is important process in pre-processing, and its precision influenced the effect of image fusion directly.In medical image processing fields, the method based on rigid registration is relatively mature now; but for it is non-rigid of human tissues and organs, so in many cases we can't process the medical image based on rigid registration. Because many properties of deformation in medical are non-rigid or non-linear, to meet the need of clinic, we must use the non-rigid registration. So multimodal medical image registration becomes a hot and difficult problem in today's image processing fields. This paper mainly studies non-rigid registration in multimodal medical image, the algorithm related and realization.For the problem of which point to start and the lack of restrictions in searching range in conventional mean pyramid structure search algorithm. Improved LOG operators would be used to extracted edge images, and build edge Pyramid images. A fast image matching algorithm would be constructed based on improved edge registration method; first match with low- resolution edge images, and then it can be implemented stepwise up to the full-resolution image matching. The results showed that the algorithm can reduce calculation amount and improve the efficiency distinctly. It reduces about 30% time than traditional medical edge registration.First, introduced the background of medical image registration, the content and field to research, the development and so on; then summed up the technology of medical image registration, included pretreatment, the classified of registration, the optimized algorithm of registration and the basic methods; and then based on the pyramid algorithm of wavelet analysis, the space change of unitary mutual information and improved edge registration algorithm; finally, analyzed the most mutual information and the choices of the labeled dot, and realized the software of medical image registration based on Visual C++ 6.0 of Windows operation system.