This paper describes the development of a vision-based, nighttime navigation support system for automotive vehicles driving in semi-structured environments. The focus is on automatic reduction of glare produced by high beams from incoming vehicles and on generation of virtual lanes that are overlaid on the image where actual road markings should exist. To accomplish these goals, we develop video-based robust techniques that produce enhanced road images in real-time. The proposed techniques are validated by experiments with real video streams, which demonstrates the system's performance and accuracy under real conditions.