CURVATURE DRIVEN DIFFUSION BASED MEDICAL IMAGE REGISTRATION METHODS

Problem statement: This paper introduces several ecient medical image registration methods. Approach: Curvature driven diusion based techniques, in particular, Perona{Malik, anisotropic diusion, mean curvature motion (MCM), ane invariant MCM (AIMCM) are employed as a regularization term in the optimal control formulation of the image registration problem. Existing optimization problem is solved using gradient-descent based optimization method. Results and conclusion: Computational results indicate that all these regularization terms produce similarly good registration quality but that the cost associated with the AIMCM approach is, on the average, less than that for the others. Exper- imental results indicate that curvature driven diusion based techniques are ecient, robust and quite fast in the nonrigid registration of medical images.