Cellular automata-based optical flow computation for "just-in-time" applications

Real-world tasks often require real-time performances. However, in many practical cases, "just-in-time" responses are sufficient. This means that systems should be efficient enough to operate on-line and to be usable in reactive systems, while being robust enough for the specific task they are performing. This paper illustrates a new "just-in-time" technique for feature-based optical flow computation on a cellular automata paradigm and, as a case study, its implementation on a special-purpose architecture for cellular automata. Feature extraction is performed by means of a simple geometrical coding based on the focal morphology of edges, which allows its description in terms of the cellular automata paradigm and reduces its temporal complexity. The experimental results demonstrate that the algorithm performs well both in controlled and uncontrolled environments.