An Image Registration Model in Electron Backscatter Diffraction

Recently, variational methods were successfully applied for computing the optical flow in gray and RGB-valued image sequences. A crucial assumption these models is that pixel-values do not change under transformations. Nowadays, modern acquisition techniques, such as electron backscatter diffraction (EBSD) used materials science, can capture images with values nonlinear spaces. Here, belong to quotient space ${SO}(3)/ \mathcal S$ of special orthogonal group modulo discrete symmetry crystal. For data, remain unchanged transformations appears be no longer valid. Hence, we propose a model determining ${SO}(3)/\mathcal S$-valued sequences, taking into account dependence on transformation. More precisely, data transformed according rotation part polar decomposition Jacobian To nonsmooth without obtaining so-called staircasing effects, using total generalized variation, prior. Then, prove existence minimizer our explain how it discretized minimized by primal-dual algorithm. Numerical examples illustrate performance method.