An image processing apparatus includes a face detection processing unit that reduces an image of a frame to be processed to a first level among several reduction levels to generate a reduced image of the frame to be processed, with one of frames included in a moving image as the frame to be processed, and compares the reduced image generated by the reducing unit and the learning data to extract a face image from the reduced image. When the extraction of the face image from the frame to be processed is ended, the face detection processing unit updates the frame to be processed to a next frame subsequent to the frame to be processed and generates a reduced image that is reduced to another level other than reduction levels adjoining the first level.