Discrete nD arrays that are constructed to have strong auto- and cross-correlation properties have found wide application in various forms of active image acquisition as well as for embedding in image data for  digital watermarking.  Their design can be often be simplified by reconstructing nD arrays from lower dimensional projected views, as projection preserves these correlation properties. In compressed sensing, a disordered pattern of sparse samples, with a density below the Nyquist rate, is used to accelerate image acquisition. Here I review some discrete image sampling schemes and examine the use of arrays with structured correlations to guide sparse sampling strategies.