In this thesis several existing approaches to Still Image Saliency are evaluated and a new approach to Motion Novelty Scoring is proposed. The purpose of Still Image Saliency is to identify interesting regions in a static image while Motion Novelty Scoring copes with the problem of detecting novel, interesting events in video sequences based on the trajectories of the objects that are moving in the scene. The results of both methods are combined in order to refine the Region of Interest chosen by a system; e.g., a mobile robot or a surveillance system. One of the key challenges in Motion Novelty Scoring, on which this thesis focusses, is to appropriately describe currently observed motion in order to enable the comparison to previously observed motion. The proposed approach tackles this challenge by dividing an object's recorded trajectory into linear and curved motion segments resulting in a so-called Motion Descriptor. To those Motion Descriptors a habituation function is applied to model the system's habituation to the motion represented by the corresponding Motion Descriptor. The Novelty Score is then directly derived from this habituation value.