بهبود قطعه بندي بافت هاي پويا و كاربرد آن در مكان يابي و قطعه بندي تصاوير MR قلب

چكيده انگليسي :

Dynamic Texture Segmentation and its Application for Heart Localization and Segmentation in Cardiac MRI Fateme Jafari Nedoushan f jafarinedoushan@ec iut ac ir Date of Submission 2011 09 21 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor Saeid Sadri sadri@cc iut ac ir Rasoul Amirfattahi Fattahi@cc iut ac ir Abstract Temporal or dynamic textures are video sequences that are spatially repetitive and temporallystationary Dynamic textures are temporal analogs of the more common spatial texture They are a family ofvisual phenomena where the texture elements or the spatially repeating elements undergo stochasticmotions that are statistically similar Dynamic textures appear in a vast spectrum of videos ranging fromsequences of moving water foliage smoke and clouds to sequences of swarms of birds robots and evenhumans in crowds Also the applications concerning such video sequences are significant and numerous including surveillance e g monitoring traffic or crowds detection of the onset of emergencies e g outbreak of fires and foreground and background separation e g the transfer of a dynamic texture fromone environment to another or simply dynamic texture removal The study of dynamic textures posesnumerous challenges especially for traditional motion models that fail to capture the stochastic nature ofdynamic textures Despite their importance the study of dynamic textures has just recently attracted theattention of the computer vision community After modeling dynamic textures it can be used for dynamictexture synthesis compression categorization and segmentation The main goal of this thesis is dynamictexture segmentation For dynamic texture segmentation first using a learning process model parametersare extracted then a region based segmentation technique using level set methods and generalization ofChan Vese model in time is designed This technique can segment dynamic texture videos into regions withdifferent spatiotemporal statistics In this method the spatial statistics of a dynamic texture is modeled witha set of texture descriptors whose temporal evolution is governed by an AutoRegressive AR model Alsodifferent feature vectors were investigated and experimental results showed that directional WalshHadamard transform has a high performance to extract texture features As an application we used dynamictexture techniques for medical issue segmentation Due to characteristics of MR images and to heartanatomy the problem of heart localization and segmentation in MRI is still open Because the heart is theonly moving organ in the images and other organs like chest and liver are static heart can be considered asa dynamic texture Modeling heart as a dynamic texture can be used for heart localization and epicardiumand endocardium segmentation A fast and robust algorithm is proposed for heart localization and dynamictexture segmentation techniques are used for epicardial segmentation Also we proposed a dynamic texturebased method for endocardium segmentation Experimental results are compared to manually segmentedresults and evaluated by numerical measure The correct localization rate is 100 and the proposed methodfor epicardial segmentation achieved encouraging results accuracy rate 0 98 Keywords Dynamic texture segmentation AR model Directional Walsh Hadamard LinearDynamic Systems Cardiac MRI endocardium segmentation

جعفري ندوشن، فاطمه

بهبود قطعه بندي بافت هاي پويا و كاربرد آن در مكان يابي و قطعه بندي تصاوير MR قلب

قطعه بندي مرز داخلي فلب , مدل AR , تبديل والش-هادامارد جهتي , سيستم هاي خطي پويا