No label defined (Q4492773)

In the Euclidean plane with a given coordinate system, the union \(F\) of a finite collection of arcs, which intersect only at end points and which are either the graphs of continuous functions or line intervals parallel to \(y\)-axis, is called a \textit{form.} A realistic model of such a plane is a page with two normal edges in the role of a coordinate system and with a piece of writing in the role of a form. When a form, being also the graph of a multi-valued function, has all its values connected (one point or an interval) it is called \textit{simple.} Then, ordered sequences \(\Lambda =(\Lambda_i), i=1,...,k\) of maximal simple forms are selected and the pair \((F,\Lambda)\) is the \textit{structure} of a form. Call the number \(k\) the \textit{height} of a form. Then, two forms \((F,\Lambda)\) and \((F',\Lambda')\) having the same height are \textit{equivalent} if there exists a homeomorphism \(h\: F\to F'\) which preserves the structure - mapping each simple form of \(\Lambda \) upon a simple one of \(\Lambda_i'.\) With this idea of equivalence as a classification criterion a ground is obtained for selecting the properties shared by equivalent forms as being \textit{form invariants.} Those points which are necessarily included as ends in each possible representation of a form as a collection of arcs are called \textit{nodes.} This paper includes a detailed treatment of node points and gives some related ideas of implementation in the case of machine character recognition. Note also that this synthetic approach could be of a larger interest being a precise mathematical conceptualization of a part of the world of our visual perception.