Theoretic And Test Research On Medical Image Processing Based On Partial Differential Equations

Recently, the Partial Differential Equation(PDE) theory in image processing is animportant progress in this field, and PDE is widely-used for digital image processing andcomputer vision. The traditional problems in medical image denoising and segmentationhave been solved by using the Anisotropic Diffusion Equation and geometric activecontours based on PDE. This dissertation focuses on improving the approach ofAnisotropic Diffusion Equation and Geometric Active Contours in some ways, andapplying the PDE to the analysis and processing of muti-photon excited fluorescenceeonfocal microscopy medical images.Firstly, based on the traditional model coupled with shock filters and anisotropicdiffusion, a new improved shock filters term and a new fidelity term, which is producedadaptive magnitude from image structural information, are proposed for the medical imageenhancement and denoising. Experimental results show that the method can not only wellrestrain noise but also enhance edge and keep much more structural information anddetailed information of a medical image and the effectiveness of our method is better.After analyzing the principles of the Bilateral filtering model, with the same theory ofand the different manifestation from the PDE model, a new gray-natural logarithm ratiorange filter kernel, which is produced adaptive magnitude from image gray distinctioninformation, has been pointed out. The new method can not only well restrain noises butalso keep much more weak edges and details of the image. Experimental results show thatbetter effectiveness could be achieved in medical image denoising with our method. And anew gray-level value natural logarithm ratio Neighborhood filters is extended with a newproposition by modifying the normalization factor. Apart from that, it can preserve theoriginal color transition of color images at the same time.The basic theory of method noise is introduced to analyze the diffusion equation andthe Total variation model in the paper. A model with parameter automatic adaptation, whichcan automatically adapt to the texture and detailed information of the medical image, isproposed for the Total Variation model. The new adaptive variation scheme can preservemore texture and small scale details encoded in medical image features in the denoisingprocesses.Through the analysis of current Geometric Active Contours, a LTP-based textureimage rapid segmentation using C-V model is proposed. In the method of the rapid Geometric Active Contours without re-initialization, the LTP operator is applied to extracttexture features, and using the LTP operator can make the texture discrimination excellentand meet the need of low complexity for texture segmentation. The experiments show thatthe proposed method is accurate and fast.A new center location method based on fuzzy mathematics is presented, and the levelset method is applied to the segmentation task. With this method, the center of karyon canbe easily found even with different sizes in one muti-photon microscopy cell image. Asshown in the result, this center location method is proved to be effective and robust forcapturing the structural information of the karyon.According to the pathological characteristics of skin hypertrophic scar tissue, texturefeatures on the edge direction, density distribution and geometric morphology are extractedfrom two-photon excited microscopic images of scar tissue and used to distinguish thenormal tissue from scar tissue. Analysis algorithms of PDE-based are applied to analyze theextracted texture features. These results indicate that texture features and analytic methodsare effective in distinguishing the normal tissue from scar tissue and accurately fixing ontheir border. And it can estimate the curative effect according to the analysis of nonlinearspectral images based on PDE.Finally, two-photon excited fluorescence based on confocal laser scanning system isused to study the cellular localization of 5-ALA induced Ppâ…¨in DHL cells, andfluorescence probes of Rhodamine123, DioC₆(3) and LysoTracker Green are used toindicate the cellular localization of 5-ALA induced Ppâ…¨with a double-labeling method,level set method is carried out to detect the image edges, extract and segment theinformation of fluorescence images. Image Localization Analysis for Two-photon excitedfluorescence can be regarded as a useful method for quantitative or qualitative cellularlocalization analysis of 5-ALA induced Ppâ…¨. And the new technology of medical imageprocessing can play an important role in applying the technology of muti-photon excitedfluorescence confocal microscopy imaging to medical diagnose and other important fields.