We proposed a visual attitude control for an outdoor autonomous mobile robot. The attitude control method uses a virtual barycenter of the quadrangle that is constructed from four feature points in camera image. By tracking four feature points using optical flow at all time, the quadrangle is updated using the same feature points, and its virtual barycenter is calculated. The quadrangle maintains the similar figure during a robot's movement. Therefore, its virtual barycenter stays at the same position when the mobile robot moves straight. On the other hand, the position of the virtual barycenter varies in proportion to the movement in the horizontal direction of the mobile robot. We demonstrated that the autonomous mobile robot equipped with this visual attitude control was able to move exactly between way points. The experimental results proved that the visual attitude control is useful for mobile robots.