Research And Implementation Of Medical Image Segmentation And Registration Algorithm

Image segmentation and registration play an important role in the image analysisand processing. Especially in medical image analysis and processing, correctsegmentation and registration are needed in the clinical treatment. Although it hasbeen many years since these problems were put forward, no prefect methods have beengiven after so many algorithms’ put out. So for the medical image segmentation andregistration algorithm research is still the focus in the field of medical image processing.This thesis mainly discusses the medical image segmentation algorithms and theregistration algorithms, and gives the improvement.Image segmentation will different image content apart or find the region of interest.When used in medical image process, it means tissue segmentation and diseased tissueidentification. In this thesis, we will mainly talk about the former and give theimprovement. We will use the level set method in the algorithms. Using the method, wewill get a good result of brain segmentation with intensity inhomogeneity. Thealgorithm is improved, and we make the algorithm convergence correctness andconvergence rate of the contradictions have improved.Image registration can correct the different positions, different rotations anddifferent deformations of the same content in the different images. Medical imageregistration contains rigid registration and non-rigid registration. This thesis explains thealgorithm framework and framework contents. Using the algorithm introduced, rigidregistration is completed for the different modes of medical image. In our registrationalgorithm, several methods on different similarity measures and optimization algorithmsis used. In addition, image registration is the premise and necessary stage of the imagefusion.Using the non-rigid registration algorithms, the local deformations can berecorrected. Non-rigid registration is more applicable to the medical image than othermethods. This thesis mainly explains the registration algorithm based on B-splinefree-form deformation and does some improvements in the medical non-rigidregistration. This method can meet the requirements of the ness on the parameters of global deformation. And the topological structures of the image contents won’t changein the registration. This method based on free-form deformation uses the grids on theimage to complete the global deformation. We combine the level set function and thefront method. After improved, the method can fixed the imperfect and improve thegenerality use.