Some embodiments involve obtaining access to a number of omnidirectional images which each represent a different viewpoint relative to the scene, and to metadata which is indicative of a physical relation between the respective viewpoint and a physical object present in the scene. Viewer data may be obtained which is indicative of a current physical relation between a viewpoint of the user and a same type of physical object in a current viewing environment of the user. An omnidirectional image may then be selected for display by comparing the viewer data to the metadata of the omnidirectional images.
In other embodiments, an omnidirectional image may have been selected, e.g., from a number of omnidirectional images or by being the only available image, and the user may be instructed to assume a viewpoint in which the physical relation to the physical object in the user's current viewing environment corresponds to the physical relation as indicated by the selected image's metadata.