Image processing on linear transputer arrays

This paper examines the computation of pixel level image processing algorithms on a linear array of processors. Occam algebra is used to show how convolution type image processing algorithms, which are fundamentally two dimensional in character can be computed on linear arrays. The paper concludes by considering the hardware/software tradeoffs that are possible for arrays implemented using transputer chips and compares the results with other parallel image processing architectures. The principle advantage of the proposed architecture is that it reduces the necessity for a frame store, and leads to very low values of data latency.