About this Issue THE FIRST tWO papers of this issue are concerned with the character recognition: handwritten numbers and letters, and printed chinese...
About this Issue THE FIRST tWO papers of this issue are concerned with the character recognition: handwritten numbers and letters, and printed chinese characters, respectively. The papers present contrasting viewpoints : Powers analyzes the actual detailed movement of the line as the basis for his features, whereas Wang and Shiau use generalized features not necesSarily related to a directly "pictorial" analysis of the character. In the third paper we are again dealing with visual patterns, but this time the problem is quite different. Ulstad is trying to compare two pictures taken of the same place at different times, to detect any differences. Next we turn to more abstract theory in pattern recognition. Kittler and Young discuss the"Karhunen-Loeve" expansion for selecting features. This analysis is used to "transform representation vectors into a new coordinate system in which the coordinate coefficients are mutually uncorrelated". Shimura and Imai, in a mathematically analogous method, investigate a nonsupervised learning method for pattern classification. Both articles present some experimental results of applying their methods. The final two papers of this issue bring to bear other disciplines on pattern recognition, namely linguistic analysis, and "fuzzy" languages. In the paper by Joshi, "some specific aspects of language structure are examined in detail in the context of pattern analysis". It is hoped that the paper "would be of general interest to experts in pattern analysis and would possibly encourage them to look at the full range of techniques and problems in linguistics and to see what extent they are relevant to pattern analysis". Thomason considers situations in which "a pattern grammar should allow for the development of strings by less-than-perfect matching of productions, with the weight of a production reflecting the degree of matching achieved". He goes on to note that "since fuzziness is inherent in many classification problems, the emerging theory of fuzzy languages has application in the handling of imperfect patterns within the frame work of formal language theory". Both papers present lucid descriptions of their approaches, attesting to the expertness of their respective authors in language usage. Also in this issue we have included a glossary, compiled by Haralick. As Pattern Recognition evolves into a major science in its own right, it must also develop a general concensus of the technical use of terms. I hope that this glossary may stimulate others to contribute thoughtful ideas and suggestions in this area, and that it will help to clarify terms used in pattern recognition. ROBERT S. LEDLEY