After immobilization, patients show impaired postural control and increased risk of falling. Therefore, loss of balance control should already be counteracted during immobilization. Previously, studies have demonstrated that both motor imagery (MI) and action observation (AO) can improve motor performance. The current study elaborated how the brain is activated during imagination and observation of different postural tasks to provide recommendations about the conception of non-physical balance training. For this purpose, participants were tested in a within-subject design in an fMRI-scanner in three different conditions: (a) AO + MI, (b) AO, and (c) MI. In (a) participants were instructed to imagine themselves as the person pictured in the video whereas in (b) they were instructed simply to watch the video. In (c) subjects closed their eyes and kinesthetically imagined the task displayed in the video. Two tasks were evaluated in each condition: (i) static standing balance and (ii) dynamic standing balance (medio-lateral perturbation). In all conditions the start of a new trial was indicated every 2 sec by a sound. During AO + MI of the dynamic task, participants activated motor centers including the putamen, cerebellum, supplementary motor area, premotor cortices (PMv/d) and primary motor cortex (M1). MI showed a similar pattern but no activity in M1 and PMv/d. In the SMA and cerebellum, activity was generally higher in the dynamic than in the static condition. AO did not significantly activate any of these brain areas. Our results showed that (I) mainly AO + MI, but also MI, activate brain regions important for balance control; (II) participants display higher levels of brain activation in the more demanding balance task; (III) there is a significant difference between AO + MI and AO. Consequently, best training effects should be expected when participants apply MI during AO (AO + MI) of challenging postural tasks.