In this paper, we propose a reliable method for view-based navigation of mobile robots fully improved in positional accuracy by using feature-points extracted by SURF, and it is verified from the navigation experiments of them. In experiments conducted in an indoor corridor with a robot for comparing the proposed method to conventional one, the positional precision was obtained in centimeter-order of within 10.0 [cm]. It suggests that the proposed method is applied to the view-based navigation for robots in such narrow areas as obstacle avoidance and corridors.