An image processing device includes a subject detecting unit detecting a subject on a material image, a working unit performing a working process on the material image such that a region of the subject puts in the block, an image generating unit clipping a subject-including region of the material image so as to have the same form and size as the block and generating a subject-including clipped image and a replacing unit setting one clipped image which is the most similar to the block in terms of color-related characteristics as the replacement image for each block of the production target image and replacing the block by the replacement image to generate a photo-mosaic image.