An image processing method, including extracting shot information from metadata, wherein the shot information is used to classify a series of two-dimensional images into predetermined units; if it is determined, by using shot type information included in the shot information, that frames classified as a predetermined shot can be reproduced as a three-dimensional image, extracting background depth information from the metadata, wherein the background depth information is about a background of the frames classified as the predetermined shot; generating a depth map about the background of the frames by using the background depth information; if the frames classified as the predetermined shot comprise object, extracting object depth information about the object from the metadata; and generating a depth map about the object by using the object depth information.