To provide an image processor which can inspect an object by automatically deciding the most characteristic area in a picked-up image.

The image processor includes a reference image area retrieval part 12 which extracts the part with the largest dispersion as a reference image from the picked-up image of good quality and a comparative inspection part 15 which compares the reference image extracted by the reference image area retrieval part 12 with a picked-up image of an object and discriminates the object or decides its quality. Since the comparative inspection part 15 compares the reference image of the characteristic part extracted by the reference image area retrieval part 12 with the picked-up image of the object and discriminates the object or decide its quality, the inspection efficiency and inspection precision can be improved.