In an image processing apparatus, minimum width and height and maximum width and height of an object to be detected are calculated on the basis of photographing conditions, an object detection range and a size of the object to be detected and an image reduction coefficient is set on the basis of the calculated minimum width and height and maximum width and heights whereby detection can be achieved while keeping the detection accuracy of image processing intact and an intruding person can be detected at a high speed at a necessarily lowest processing speed.