This paper deals with the autonomous navigation scheme for a mobile robot in indoor environment using an upward-looking camera and sonar sensors. Corner and lamp features are extracted from the sequential ceiling images, and these features are used as landmarks in the SLAM (simultaneous localization and mapping) process. Combining lamp information with the conventional corner feature-based approach provides accurate pose estimation, since lamp features are robustly detected and associated in most indoor environments. The extracted features are used in the EKF (extended Kalman filter) to estimate both robot pose and feature positions. Based on the pose estimation from the SLAM process, autonomous exploration is achieved by applying driving gains to exploration nodes. The sonar sensors are adopted to detect most obstacles including glasses and black surfaces. The proposed scheme is a low-cost solution to autonomous mobile robot navigation since it can be implemented with a web camera and a small number of sonar sensors. Experimental results show that the proposed scheme works successfully in real environments.