Sensorimotor processing specifically impacts mental body representations. In particular, deteriorated somatosensory input (as after complete spinal cord injury) increases the relative weight of visual aspects of body parts’ representations, leading to aberrancies in how images of body parts are mentally manipulated (e.g. mental rotation). This suggests that a sensorimotor or visual reference frame, respectively, can be relatively dominant in local (hands) versus global (full-body) bodily representations. On this basis, we hypothesized that the recruitment of a specific reference frame could be reflected in the activation of sensorimotor versus visual brain networks. To this aim, we directly compared the brain activity associated with mental rotation of hands versus full-bodies. Mental rotation of hands recruited more strongly the supplementary motor area, premotor cortex, and secondary somatosensory cortex. Conversely, mental rotation of full-bodies determined stronger activity in temporo-occipital regions, including the functionally-localized extrastriate body area. These results support that (1) sensorimotor and visual frames of reference are used to represent the body, (2) two distinct brain networks encode local or global bodily representations, and (3) the extrastriate body area is a multimodal region involved in body processing both at the perceptual and representational level.