In a device for combining a plurality of images having a same photograph range, a position shift between the images having the same photograph range is detected, and the position shift of the images is corrected, and after the position shift is corrected, similarity of the respective portions between the images is evaluated, and the image combining method is changed according to the similarity evaluation. Therefore it is possible to create an image having no position shift due to the shaking of a camera and object blurring due to the movement of the object. Moreover, whether the images are combined or not is determined according to the similarity, hence the processing load is reduced and an image having no object blurring can be created.