An image extracting apparatus is provided which can automatically finish a captured image of a person to provide an easy-to-view picture. It includes an image input unit supplied with a captured color image of the person to output it as digital data, a flesh-color region extraction unit supplied with the digital image data to detect a flesh-color region in the image, an object-image detection unit to detect an object image from the detected flesh-color region, and a trimming unit to trim the detected object image. On the assumption that a region extending from the top end of a certificate picture to the head top of a person is an overhead region A, a region extending from the head top to the jaw is a face region B and a region extending from the jaw to the bottom end of the certificate picture is a chest region C, the trimming unit trims the image so that the dimensional ratio between these regions A, B and C is 1:0.4 to 0.8:0.1 to 0.26.