Showing papers on "Adjacency list published in 1973"
01 Oct 1973
TL;DR: A methodology to analyze the pictorial information represented in a map-like image, and further name the objects most likely present in the scene based on the universe known to the system is developed, which is flexible enough to recognize incomplete objects, and has learning capabilities through adding and modifying the rules of the system.
Abstract: Thesis. A methodology to analyze the pictorial information represented in a map-like image, and further name the objects most likely present in the scene based on the universe known to the system is developed. In contrast to conventional string language sentences which are formed by concatenating the primitives (symbols) in a string according to the language grammar, the pictorial sentences are basically formed of primitives which are located in a twodimensional (or three-dimensional) space with much richer relational properties than simple adjacency. The possibility of describing the structural information of pictures as syntactical rules, employing a rich class of relations, was examined. The gnaphs are known to be extremely convenient and flexible to represent this kind of information. By expressing the syntactical properties of one class of pictures as a collection of graphs, it was shown that algorithms can be easily developed to parse objects or collections of objects which belong to this class. Further, this methodology is flexible enough to recognize incomplete objects, and has learning capabilities through adding and modifying the rules of the system. In addition, it was shown that a complete description of an object as a collection of gnaphs can be easily modified (rotated) tomore » enable the system to recognize different projections of the object. (auth)« less
3 citations