Nachricht

Whereas for human vision and visual attention the relevance of shape features has already been researched on a large scale, evaluations of possible features for artificial attention have been done rarely. Particularly, it is unclear if, besides established shape features inspired by human vision, others might be suitable for attention models as well. In this thesis, both types of spatial features are tested, evaluated and optimized. This also contains the implementation of new features in the region-based saliency model.