Frameless computing for spatial-temporal events

The detection of spatial-temporal events is a difficult task in machine vision and it is usually difficult to be handled efficiently with current algorithms and devices. There are many examples in nature, like looming detection or detection of moving objects with given speed and trajectory, which shows that human vision system can solve this extremely difficult task with extremely low power consumption. In this article we show examples how cellular neural networks can be used to detect spatial-temporal events. The detections are done by using continuous dynamics without cutting the input flow into frames. We can observe similar structures and functions (the detection of continuous input-flows with continuous dynamics) in the retina, which performs well and efficiently in image processing tasks.