To increase traffic safety, automotive companies and governments have developed various kinds of technologies to support the needs of reducing traffic accidents during the past decades. To achieve this goal, vehicular ad hoc network (VANET) technologies have been getting the spotlight recently and extensive researches using VANET technologies have been carried out. Among them, decentralized congestion control (DCC) of periodic beacon broadcast to support cooperative collision warning (CCW) applications is one of the most critical and challenging issues in VANETs. In order to realize DCC, however, the accurate vehicle density estimation should be preceded because vehicles adjust their congestion control parameters according to the estimated vehicle density. In this paper, we present the existing vehicle density estimation techniques in vehicular communications and point out their limitations and technical challenges for DCC in vehicle-to-vehicle (V2V) safety communications. The question about how to solve these problems is also discussed in this paper.