In QoS networks, Connection Admission Control (CAC) is required to ensure that network resources are used efficiently. Very often CAC algorithms attempt to estimate available network resources (e.g., bandwidth, buffer space) to make admission decisions. In the case of real time video applications, transmission delay and packet loss rate constraints are crucial QoS parameters. However, the mapping between these parameters and available network resources is not so easy. In this paper, we present and analyze an adaptive admission control and traffic adjustment algorithm for real time video transmission, which is based on the measurement of end-to-end delay probability distribution and the user-defined QoS level profiles. As the network status varies, the algorithm attempts to improve or degrade the QoS levels of existing calls within the acceptable range. Simulation results demonstrate that the proposed algorithm performs well in terms of decreasing the call blocking rate while utilizing the network bandwidth effectively.