This thesis develops a new iterative method for a class of convex constrained optimization problems, which occur inter alia in imaging models based on variational approaches. The new method overcomes problems of established optimization methods, such as step size restrictions and resulting slow convergence. Convergence properties are derived, based on monotone operator theory. Numerical experiments, primarily on the application of real-time motion segmentation, for which suitable models are developed, show the performance of the new method.
«
This thesis develops a new iterative method for a class of convex constrained optimization problems, which occur inter alia in imaging models based on variational approaches. The new method overcomes problems of established optimization methods, such as step size restrictions and resulting slow convergence. Convergence properties are derived, based on monotone operator theory. Numerical experiments, primarily on the application of real-time motion segmentation, for which suitable models are deve...
»