To improve accuracy for separating a dot region and a character region from binary image data where the dot region and the character region coexist.

A black hat processing is performed on the binary image data consisting of a black pixel and a white pixel. Thus, a background region formed of the white pixel sandwiched between the black pixels is extracted, and a closing processing is performed on background region image data including the extracted background region. Thus, a mask pixel region showing the dot region is extracted and the character region which is erroneously detected as the dot region is deleted from the extracted mask pixel region. Thus, the mask pixel region is corrected and the dot region and the character region are separated from the binary image data based on the corrected mask pixel region. When the mask pixel region is corrected, an area of the mask pixel region is compared with a minimum area which is previously set. When the area of the mask pixel region is smaller than the minimum area, the mask pixel region is determined to be the character region and the character region is deleted from the mask pixel region.