An image processing device includes a processor including hardware. The processor is configured to: acquire an image data including an annular side observation image and a direct view observation image forming a circle formed from a center of the side observation image to an inner circle; in a case of normal observation in which electronic zooming processing is not performed, generate a first display image including a boundary portion between the side observation image and the direct view observation image in a display area; and in a case of magnified observation in which the electronic zooming processing is performed, generate a second display image in which a boundary portion between the side observation image and the direct view observation image is shielded by a mask.