An image processing device comprising a representative parallax acquisition unit, a scene separation unit to separate the stereoscopic video into multiple scenes when a parallax width does not comply with an allowable parallax width, a parallax adjustment unit to decide whether a scene parallax width complies with the allowable parallax width, and uniformly adjust the representative parallaxes for the respective stereoscopic image frames constituting the scene such that the scene parallax width complies with the allowable parallax width, the scene parallax width being defined by a maximum value and a minimum value of the representative parallaxes for the respective stereoscopic image frames constituting the scene, and an output unit, wherein the representative parallaxes for the respective stereoscopic image frames include a statistical operation value to be calculated based on parallaxes that, of parallaxes for the stereoscopic image frames, meet a predetermined condition.