A three-dimensional-object can be effectively detected. A pair of image capture devices capture a three-dimensional-object and calculate disparity component data of subdivided-image regions, respectively. On a basis of the disparity component data, gray scale values indicative of distances from the image capture device are calculated and a gray scale image in which each region has its corresponding gray scale value is generated. A model of the three-dimensional-object is defined and correlation values is calculated to show the degree of a similarity between the model and the image subdivided regions in the gray scale image. The model is a two-dimensional image with a shaping characteristic when the three-dimensional object is viewed from positions of the image capture devices while each subdivided region of the two-dimensional image has a gray scale value indicative of a distance from the image capture device at a portion corresponding to the three-dimensional object. The correlation values are calculated on a basis of gray scale values of the model and those of the image region in the gray scale image. The model and an image region with the highest correlation value are detected in the gray scale image, so that a three-dimensional image is detected.