An image processing apparatus includes an analysis processing unit that quantizes an input image in two levels by replacing an area of the input image where a feature of a subject appears with a first pixel value and replacing an area of the input image where no feature of the subject appears with a second pixel value when the input image is acquired; a storing unit that stores dictionary data including a feature of a plurality of face images quantized by the analysis processing unit; and a face detection processing unit that performs face detection on a detection target image by using the dictionary data read from the storing unit, the detection target data being quantized by the analysis processing unit.