An image processing device includes a stereoscopic image supply unit configured to supply a stereoscopic moving image in which stereoscopic image pairs each including two images of an object viewed from mutually different viewpoints are arranged in a temporal order, a synthetic image generation unit configured to generate a synthetic image in the temporal order by performing a predetermined operation on luminances of the two images, and an output image selection unit configured to sequentially select and output any one of the two images and the synthetic image in the temporal order so that a result of a time integral during a predetermine period of an output thereof offsets any one of the two images.