MAP: Multi-angled parallelism for feature extraction from topographical maps

Abstract MAP (multi-angled parallelism) is proposed as a unification of conventional erosion-dilation operations for binary images and higher order feature representation where each pixel is composed of multiple directional elements. A “twisting operation” is also proposed which is able to have any direction of propagation and any directional width using 8-neighbor operations. The MAP concept can be effectively and precisely achieved using the twisting operation. By combining MAP with vibrational operations such as erosion-dilation and dilation-erosion, a set of MAP operations to extract linear features is defined. These MAP operations have been applied successfully to the extraction of straight lines, sets of points, buildings, hatched areas, roads and character regions of topographical maps. Such features are difficult to extract using conventional parallel operations.