A method forms a region image extracting a region of a physical object which generates three-dimensional model information from a video image of physical space. Then, from position and orientation information for the physical object and from the region image, a primitive virtual object of a size to encompass the region image is generated. From the primitive virtual object a virtual object having a shape according to the region image is generated, and the three-dimensional model information is generated as three-dimensional model information representing the physical object.