Research On Medical Image Segmentation And Registration

Image segmentation is one of the key problems in image analysis and processing. Classical image segmentation techniques cannot satisfy the requirements of complex image segmentation because of the limitations itself, under the circumstances, model-driven image segmentation techniques become a focus of widespread concern. Geometric Active Contour Model is capable of handling changes in the topology of the evolving contour, while it is a difficult problem for parameter active contour mode. And the using of the level set method, which has greatly promoted the development of the geometric active contour models.Traditional level set method of C-V lack of controlling the local feature. In order to eliminate C-V method's defects, a novel segmentation model based on exponential boundary gradient speeding term is proposed. By incorporating the local image information into the proposed model, the gray uneven image can be efficiently segmented in less iteration. And the energy penalty term eliminates the time-consuming re-initialization process of C-V model. In addition, an evolution termination criterion based on the curve length change is proposed to ensure that the evolving curve can automatically stop on the true boundaries of the objects. Large numbers of experiments indicate that the proposed model can not only selectively speed the segmentation of specific objects, but also improve the segmentation accuracy of objects with weak boundaries.Medical imaging has aided clinical diagnosis and clinical applications by providing multi-modality medical images. And different imaging equipments provide relatively complementary information of the same anatomy location of patient. In order to get more comprehensive and reliable information to facilitate clinical diagnosis, it has been commonly desirable to combine the information provided by different modality images in the clinical application. The aim of medical image registration is to normalize images of different imaging time or different modality into a common coordinate. It has been significance for clinical comparison and analysis to make the individual information represent simultaneously in the same coordinate.To address the disadvantages of traditional image registration, a new method for image registration is proposed that combines hybrid genetic algorithm with wavelet multi-resolution analysis strategy. In this method, the mutual information is used as the similarity measure and a hybrid genetic algorithm is used as the search technique, and the parameters of genetic algorithm are adapted along with the computing of mutual information and the multi-resolution of the images. The experimental results show that this registration method could efficiently restrain the local maxima of mutual information function and the subvoxel accuracy can be achieved, which demonstrates that the algorithm is accurate, robust and efficient for image registration.