Algorithms Study Of Multi-Modality Medical Image Fusion And Target Object Segmentation And Localization

This thesis carries out the study of the medical images processing of Heavy Ion Beam 3D Treatment Planning,including medical image preprocessing,multi-modality medical image registration and fusion,the region of target segmentation.Various algorithms of the processing were studied and improved,at the same time,the experimental platform based on Visual C++ 6.0 was completed to verify the effect of algorithms.Firstly,In medical image preprocessing filed,the image filtering and the enhancement method based on wavelet transform was especially studied.The hybrid filtering-enhancement algorithms combining median filtering with soft threshold filtering-enhancement algorithm based on wavelet transform was presented,it fulfilled the purposes of denoising and enhancing the edge information.Secondly,in multi-modality medical image registration and fusion filed,the interpolation algorithm, similar measure and search algorithm of the registration were studied and improved. The medical image registration based on mutual information was adopted for the characteristic of different-modality medical image registration. Especially,optimization technique for image registration was improved,the hybrid PSO algorithm based on the PSO algorithms and simulated annealing algorithm was introduced,and the comparing experiments shows that the improved optimization algorithm is superior to the standard particle swarm algorithm.In addition,Image fusion algorithms based on airspace and wavelet transform were analyzed in this paper.The improved multi-resolution wavelet tower fusion algorithm was presented,and the algorithm fuse all information of the original image, it also enhance image edge information.Finally, in medical image segmentation filed, the research was mainly based on the deformable model algorithm of the image segmentation,and the parameters deformable model and geometry deformable model were included.The level set algorithm of geometric deformable model was improved,the improved algorithm is better in speed and segmentation result.