Introduction Research Statement.pdfVIP

Research Statement Igor Yanovsky Introduction I have a very diverse background in computational and applied mathematics. Areas of my research include variational methods, computational fluid dynamics, level set methods, multigrid methods, and information theory. I applied knowledge of these areas to solve problems in computer vision, image processing, and biomedical image analysis. In particular, my dissertation work links all aforementioned areas of computational and applied mathematics to solve image registration and segmentation problems in computer vision and medical imaging. I have been fortunate to work under the guidance of my advisor Prof. Stanley Osher and collaborate with one of the world’s best known brain imaging groups, led by Prof. Paul Thompson, in the Lab of Neuro Imaging at UCLA. For my doctoral dissertation, we proposed the Unbiased Large-Deformation Nonlinear Image Registration model, which is described in Section 1. My thesis work also includes a novel technique for segmentation of homogeneous deformation described in Section 3. Besides my dissertation work, my research included the areas of multiphase image segmentation, surface denoising, surface reconstruction, and Eulerian vortex motion. 1 Unbiased Large-Deformation Nonlinear Image Registration In this work, we propose a novel framework for constructing large deformation log-unbiased image reg- istration models that generate theoretically and intuitively correct deformation maps [8], [20]. Let I1 and I2 be two images defined on the spatial domain ?. The problem of image registration is to find a transformation g such that the deformed image I2 ? g is in some sense close to I1. In this work, we will restrict the map g to be differentiable, one-to-one, and onto. The inverse map of g is denoted as g?1 and the Jacobian matrix of g as Dg. The Jacobian map can thus be defined as the determinant of the Jacobian matrix |Dg|. The displacement field u(x) from the position x in the deformed image I2 ? g