With the increasing usage of GPS-enabled devices which can record users’ travel experiences, moving object trajectories are collected in many applications. Raw trajectory data can be of large volume but storage is limited, and direction-based compression to preserve the skeleton of a trajectory became popular recently. In addition, real-time applications and constrained resources often require online processing of incoming data instantaneously. To address this challenge, in this paper we first investigate two approaches extended from Douglas-Peucker and Greedy Deviation algorithms respectively, which are two most popular algorithms for trajectory compression. To further improve the online computational efficiency, we propose a faster approximate algorithm with error bound guarantee named Angular-Deviation. Experimental results demonstrate it can achieve low running time to suit the most constrained computation environments.