An image processing apparatus includes an image obtaining section for obtaining images from cameras on a moving object; a correspondence processing section for generating a correspondence between characteristic points extracted from a pair of images before and after the moving object moves straight; a height/orientation estimating section for estimating orientations and heights of the cameras, by estimating a homography matrix from the corresponded characteristic points, then decomposing the homography matrix, and using the decomposition result, a movement direction vector, a road surface normal vector, and a moved amount of the moving object; a projective transform section for applying projective transform to common view parts between the cameras; and a relative position estimating section for estimating a relative position of one of the cameras relative to another one of the cameras using translation for overlapping the common view parts after projective transform.