Computer vision in undergraduate education: modern embedded computing

Computer vision has historically been taught as a graduate subject since few examples of the discipline were being practiced in mainstream engineering. In recent years, the incorporation of multimedia into embedded devices has drawn some vision topics into mainstream attention. Examples of consumer products include digital video recorders, cellular phones, and automobile collision-avoidance systems. This paper describes the development of an undergraduate course that incorporates some vision topics into the larger context of embedded computing. Traditional topics, such as processor types, dynamic power management, and real-time scheduling, are taught alongside relevant vision topics, such as codecs, concurrent interfaces, and multimedia signal acquisition, storage, and rendering. In lab work, the students program hardware to operate as a digital video camera. While the primary goal for the course is to teach embedded computing, a secondary goal for the course is to entice students into graduate study in computer vision. However, a major developmental point was to justify the vision content in the context of how it serves the needs of students not opting for graduate study, as well as how the course would impact students working in other related graduate research areas.