A Context Algorithm for Pattern Recognition and Image Interpretation

One method of interpreting aerial photographs is to divide a frame of imagery into square cells, to extract recognition data from each cell, and to classify one cell at a time according to statistical decision theory. An algorithm for extracting contextual information from neighboring cells to improve cell recognition performance in such a system is presented. The algorithm was tested with a Monte Carlo technique which drew contextual information from real imagery but which simulated the as much as one half by the addition of context, and the amount of improvement was found to be largely independent of the parameters of the simulation.