This paper is concerned with the problem of an efficient region representation for a subsequent handling, mainly for geographic information processing application.
An input image (a geographic map) is assumed to be a binary one with all region boundaries coded by Freeman chain code (eight basic compass directions coded by 0 through 7 with the vertical and horizontal step of size h determined by digitization procedure). The map is divided into the horizontal parallel stripes of the height h. An intersection of every stripe and the regions of the map is a set of substripes, each of which has a trapezoidal form with the west and east sides represented by a link of chain code with one of three possible types of slope (south-north, southwest-northeast, southeast-northwest). Thus, a stripe can be uniquely coded by its ordinate, a number of the substripes belonging to it, and the abscissae and types of the west and east sides of every substripe. This coding scheme is called the Parallel Connected Stripes (PCS) by the author.
A description and pseudo-PASCAL code for the conversion algorithms from the Freeman chain code to the PCS, and from the PCS to the Freeman chain code, are presented. The algorithms for the basic set operations (intersection, union, complement) between two PCS coded maps are also given.
The author claims that PCS data structure is better suited for the efficient geographic information processing than classical raster, vectorial, and cellular representations. This claim is supported by a qualitative comparison of the aforementioned data representations. However, no comparison with the quadtree coding scheme [1], which has attracted much attention over the past several years, is given. There are no processing results for the real geographical data. It would also be desirable to provide the reader with the data on memory consumption and execution time for the presented algorithms.
The style of this paper is for the most part clear and understandable. The appendix, which contains the pseudo-PASCAL code for the algorithms, is somewhat lengthy. It is my feeling that there is no need for the detailed description of all the possible branches in the several algorithms which are similar by nature.
Despite the lack of the abovementioned data, which would make this work more comprehensive, the paper is worth reading for researchers working in the area of geographic information processing systems.