Location: MIC-Arena, Maria-Goeppert-Str. 3, 23562 Lübeck
Slice interpolation is one of the most important concepts in medical image processing, that many attempts have been made to improve it. Intensity-based interpolation and object-based interpolation are two main groups of these kinds of interpolation methods.
In this thesis, the object-based interpolation based on the modified version of the curvature registration has been proposed. Due to the nonlinear nature of image registration, forward and backward registration results are different from each other. Therefore, with the assumption of linear displacement between corresponding nodes of image grid, the modified energy functional is minimized.
This minimization is an unconstrained optimization. To solve this problem, one can use indirect method; by using the Euler-Lagrange equation from the calculus of variations, the optimization problem converted to a differential equation, that is solved by using the finite difference method. The answer of this minimization is the displacement vector of reference and template images, such that they are overlapped and the middle slice is computed. The results of this interpolation are compared with interpolation based on Demon registration.
Also, in order to optimize the amount of balanced coefficient of energy functional, the PSO metaheuristic algorithm is used. The produced results show significant visual improvement in the edges of the interpolated slices.
