This paper presents a robust roadway detection method by direction using online CCTV videos. In traffic surveillance using video/image processing, the roadway detection in the traffic scene not only improves efficiency of vehicle detection and tracking, but also reduces detection errors caused by objects peripheral to the roadway. The proposed method detects each traffic direction of the roadway separately in order to provide traffic information of both directions in a step-wise procedure. First, a roadway mask image of a traffic scene is created by accumulating moving parts in a difference image between two consequence input images. Next, a center line of the roadway is identified and used to divide the roadway mask into two parts. Roadways with different geometry types and traffic condition were studied. Experimental results show that the proposed method successfully separated the roadway directions