An intelligent image sensor based on two-dimensional cellular automata

The design of an intelligent image sensor based on two-dimensional cellular automata is presented. This intelligent sensor performs simple first-stage preprocessing of the image in parallel with the sensing operation. The sensor acts, in parallel with the sensing operation and that of thresholding, as a spatial low-pass filter on the image. A photodiode is used as a sensing element. The thresholding circuit is simply an inverter. The low-pass filtering operates to remove bright spikes, which would incorrectly appear in the thresholded image. This is performed by each pixel cell according to a local transition function, which takes into account the majority of the values of its four nearest-neighbour pixel cells. The sensor is addressable like a dynamic memory, and hence techniques such as windowing can be performed rapidly without the large data redundancy that occurs with serially addressed sensors. >