Disclosed is a method of storing scene comprising a three-dimensional image, which has been generated by converting an image input an apparatus supporting photographing of the three-dimensional image, into a recording medium. The method includes: loading, in a predetermined file type field, image type information indicating one of a single stereoscopic image and a synthetic image including a monoscopic image and a stereoscopic image; loading, in a configurative information container field, scene descriptor information indicating temporal and spatial relations between a plurality of media objects included in the three-dimensional image, object configurative information indicating object attributes of each media object of said plurality of media objects and a configurative relation between encoded streams of each of the media objects of said plurality, configurative information of the encoded streams, first metadata between the plurality of media objects, and second metadata between each media object of the plurality and the encoded stream thereof; and loading, in a media data container field, a bit stream of each media object included in the three-dimensional image, a bit stream for the first metadata, and a bit stream for the second metadata.