A stereo digital camera which generates a three-dimensional image or stereograph having one file structure, and records the three-dimensional image on a recording medium includes a stereo adapter having optical axes corresponding to parallax, an imaging lens for transferring an object image via the stereo adapter, and a single CCD pickup unit having monocular regions on which the monocular images of the object image are projected by the imaging lens. Two monocular region images from the monocular regions form one multocular stereo image, and are compressed as one image data. This image data is appended with header information which contains an item indicating that this image data is a stereo image, an item corresponding to the number of monocular images which form the stereo image, and an item associated with addresses of the first and second monocular images, and is stored in a recording medium as one file.