Generation of roadside panoramic images without obstacles

This paper presents the method for generating panoramic image(s) from plural in-vehicle camera images. Our real-world map system employing the proposed method has the following features: 1) it puts the generated panoramic images of the real-world to an existing web map, e.g., GoogleMaps, 2) partial regions in the images are classified according to the depth from the camera by tracking each partial region employing the SIFT feature, so that 3) it can obtain panoramic background images that exclude obstacles, e.g., parked vehicles. Our updatable real-world map system can offer up-to-date information to map-based systems, e.g., car navigation system or web interface any time.