The thesis concerns the hexagonal sampling of images, the processing of industrially derived images, and the design of a novel processor element that can be assembled into pipelines to effect fast, economic and reliable processing. A hexagonally sampled two dimensional image can require 13.4% fewer sampling points than a square sampled equivalent. The grid symmetry results in simpler processing operators that compute more efficiently than square grid operators. Computation savings approaching 44% arc demonstrated. New hexagonal operators arc reported including a Gaussian smoothing filter, a binary thinner, and an edge detector with comparable accuracy to that of the Sobel detector. The design of hexagonal arrays of sensors is considered. Operators requiring small local areas of support are shown to be sufficient for processing controlled lighting and industrial images. Case studies show that small features in hexagonally processed images maintain their shape better, and that processes can tolerate a lower signal to noise ratio, than that for equivalent square processed images. The modelling of small defects in surfaces has been studied in depth. The flexible programmable processor element can perform the low level local operators required for industrial image processing on both square and hexagonal grids. The element has been specified and simulated by a high level computer program. A fast communication channel allows for dynamic reprogramming by a control computer, and the video rate element can be assembled into various pipeline architectures, that may eventually be adaptively controlled.