An image processing device includes a processor including hardware. The processor is configured to: identify an image selected according to a user input operation received by an input device, from among a plurality of first images acquired by a capsule endoscope, as an identified image determine whether or not the identified image is one of a plurality of second images acquired during a reciprocating movement of the capsule endoscope and in response to the identified image being determined to be one of the plurality of second images, extract an image acquired in a first forward path or a last forward path of the reciprocating movement from among the plurality of second images.
