The major objective of this thesis is to refine the novel approach of a fixation-based segmentation by Mishra et al. called Active segmentation with fixation in oder to enable it to run on a robotic real-time system.
In contrast to classical segmentation methods where an image of a scene is divided into multiple individual regions, the fixation-based approach redefines the segmentation process by imitating the human visual system, thus it only separates the fixated region from the rest of the image.
The key to this approach is the application of the image segmentation in polar space.

The framework developed in this thesis implements several optimizations and extensions of the original approach.
The basic idea is to reduce the strong dependency of the results on the edge detection by introducing an additional optimization step - the Grab Cut algorithm.
In order to incorporate the Grab Cut algorithm, it is examined how to bridge the information gap between the given fixation point and the information needed for the Grab Cut to produce correct segmentation results.

The solution is to include the original method as an intermediate step of the new algorithm.
This yields a new and more balanced algorithm which can be used in a robotic real-time system.