Automatic detection of crowd congestion in high density crowds is a challenging problem, with substantial interest for safety and security applications. In this paper, we propose a method that can automatically identify and localize congested regions in crowded videos. Our proposed method is based on the notion that pedestrians in the congested region follow a particular behavior. Pedestrians in the congested areas cannot move freely due to space unavailability and tend to undergo lateral oscillations. In our method, we first extract trajectories by using particle advection technique and then compute oscillatory features for each trajectory. Trajectories with higher oscillation values and with less proximity are clustered, indicating the congested regions. We perform experiments on a diversity of challenging scenarios. From the experimental results, we show that our method provides precise localization of congested regions in crowd videos.