In this paper, a system for measuring flow of pedestrians using a stereo camera is proposed. The system does not need adjustment after installation and its installation environment is not restricted. The proposed method consists of two stages, i.e., estimation process and measurement process. Estimation process automatically estimates parameters that are typically adjusted manually after installation by using images and distance information obtained from a stereo camera. Measurement process measures flow of pedestrians using one of the two methods, i.e., particle filter and a method using KLT and Voronoi. The selection is performed depending on the congestion level of the pedestrians. The proposed method is evaluated through experiments for four scenes that are typical in real operation.