Improved information on land cover is required to aid our management and understanding of the environment. Remote sensing provides, commonly, the best method for obtaining geographically and temporally detailed estimates of land cover and its changes. However, the large ground area of each pixel, relative to the size of typical land cover parcels, means that pixels often represent a mix of land covers. The presence of such pixels can adversely affect the performance of traditional per-pixel classification approaches. Soft classification techniques have been developed to estimate the class composition of image pixels, but their output provides no indication of how these classes are distributed spatially within the pixel. The use of Hopfield neural network to map the spatial distribution of land cover classes using prior information of pixel composition determined from soft classification is demonstrated. An approach was adopted that used the output from a soft classification to constrain a Hopfield neural network formulated as an energy minimisation tool. Energy functions were defined for various tasks, consisting of a goal and several constraints, and the network converges to a minimum of these functions. The energy minimum represents an estimated map of the spatial distribution of land cover class components in each pixel. The Hopfield neural network design was tailored to meet specific goals, including, super-resolution target identification, land cover mapping and land cover pattern prediction. The approach was applied to synthetic data, simulated remotely sensed imagery and real Landsat Thematic Mapper imagery, with high levels of mapping accuracy and improvements over traditional techniques produced throughout. Overall, we show that the spatial resolution of satellite sensors imagery need not necessarily represent a limit to the spatial detail obtainable inland cover maps derived from such imagery. The Hopfield neural network technique is shown to be robust, efficient and simple, and results suggest that it has the potential to predict accurately land cover at the sub-pixel scale from operational satellite sensor imagery.