During the last several years, many visual communication systems that can be operated on Internet in real-time have been developed widely. These systems' target bit-rate is about several tens of kilo bit per second and supported service quality is not so good due to low capability of networks and the lack of stream processing technology. There are many available high performance real-time MPEG-1 codecs targeted to storage media applications. With them, we have implemented a desktop-based visual communication system with better visual quality operated on more wide bandwidth.
This system, however, suffers from large end-to-end delay and cannot cope with network congestion that may occur in currently available networks. This thesis suggest simplified transmission strategy satisfying various buffer constraints, clarifies the nature of end-to-end delay and proves the minimum bound of total delay empirically. Also, control methods which can be applied to unstable network environments adaptively are proposed and verified by experiments.