Despite the rapid development of cyber technologies, today we still have very limited attention and communication bandwidth to process the increasing information flow. The goal of the study is to develop a context-aware filter to match the information load with particular needs and capacities. The functions include bandwidth-resolution trade-off and user context modeling. From the empirical lab studies, it is found that the resolution of images can be reduced in order of magnitude if the viewer knows that he/she is looking for particular features. The adaptive display queue is optimized with real-time operational conditions and user's inquiry history. Instead of measuring operator's behavior directly, ubiquitous computing models are developed to anticipate user's behavior from the operational environment data. A case study of the video stream monitoring for transit security is discussed in the paper. In addition, the author addresses the future direction of coherent human-machine vision systems.