The long acquisition times (up to hours) in PET brain imaging bear a high risk of head motion, which results in artefacts like blurred images and may even lead to misinterpretation and useless data. With the increased resolution of high performance PET scanners, the influence of head movements becomes more and more relevant. Especially in the analysis of small brain structures, e.g. during ROI-analysis, head motion results in inaccuracies of quantified data. This may also influence the kinetic analysis and generate artifacts in parametric images calculated from a motion-affected image sequence. This work presents the feasibility of head motion registration using an external motion tracking system. The implementation of the multi acquisition frame method and an event-by-event method to correct PET data for motion are described. The effects of motion correction are demonstrated on the basis of phantom measurements and patient data. The influence of motion correction on parametric imaging is described in a receptor study.