An Approach to the Definition, Description, and Extraction of Structures in Binary Digital Images

A method is presented for the definition and description of structures, that is, sets of visible elements in a binary digital image. The method is designed so as to facilitate the extraction, analysis, and manipulation of structures and is based on a numerical code which synthesizes the description of an element in its 8-neighborhood. Taxonomies of features and structures are derived and discussed. The exploitation of APL typical "many-words-at-a-time" computing notation allows us to show how every defined structure may be derived and described by the use of simple techniques of parallel selection and sorting. The same techniques are used in the definition of a parallel algorithm which identifies the descriptions of the connected components and assigns them intrinsic labels by establishing relations of surroundedness. The inverse parallel algorithm, which displays the binary image corresponding to a given description, is thereafter described. The method has been tested in different practical experiments. Hints are then given on strategies which were originated using the method in practical experiments.