idea in theory within the practical area of picture processing. This paper is of3180 interest for designing techniques using global or local picture transformationsf3180 (filtering, restoration, etc.).:9L

g3181 In an earlier paper ¿1¿, the authors defined a Table Matrix L System (TMLS),3181 which generates various classes of rectangular arrays of symbols. In this 3181 paper, they introduce context dependent rules in a TMLS by allowing thei3181 replacement of a symbol to depend on any or none of its four neighbors. TheT3181 resulting system is called a TM:Ci:ALS where :Ci:A = 0,1,2,3,4 denotes thee3181 number of neighbors involved. They show that the generative capacity of ae3181 TMLS is increased by the use of context dependent rules and that the family3181 of all TM:Ci:ALSs is not closed under many operations on rectangular arrays.3181 They study a hierarchy of one-dimensional TML and TM:Ci:AL languages anda3181 compare them with known string language families generated by L systems.a3181 They also note that topics for future research include the effect ofe3181 nonterminals and coding on a TM:Ci:ALS and its grammatical inference problem.3181 The topic and results obtained are of interest to a very limited audience. :9L-3182 This paper examines the interrelationships between fuzzy relations on a group S-3182 and fuzzy subgroups on S × S. The applications of the theory of fuzzy setsu3182 include the design of robots and computer simulation. The importance of groupe3182 theory in other areas has motivated this work on fuzzy relations defined onu3182 groups.

e3182 An opening section presents some of the standard definitions and results onu3182 fuzzy sets, subgroups, and relations. Some preliminary results on fuzzye3182 relations then follow. The main theorem is this: Let A be a fuzzy subset on a 3182 group S, and let B be the strongest fuzzy relation on S that is a fuzzyt3182 relation on A. Then A is a fuzzy subgroup if and only if B is a fuzzy subgroup.n3182 After the lengproof, further results and consequences follow. Directionsu3182 for future work are also given. This theoretical work makes a good contributionn3182 to the area of fuzzy group theory, and competently extends prior work.:9Lb3186

This paper presents the application of texture segmentation techniques to a3186 particular problem in seismic data interpretation. The two textural feature 3186 extractors used in segmentation are a simple template matching scheme andr3186 grey-level run-length statistics. The derivation of a multimembership template 3186 matching algorithm is given, as is the formalization of the run-lengtht3186 algorithm.

h3186

An interesting discussion is presented on the need to modify existing algorithms3186 to meet the particular needs of this application, a situation common ini3186 pattern recognition practice. Given the title of the paper, there is not muchg3186 :2OAI” present; only three simple (and extremely low-level) heuristics arec3186 used to aid the classification of individual pixel blocks. A standardc3186 depth-first search uses these heuristics to refine decision making. No results3186 in terms of accuracy of segmentation, etc., are presented. This reviewer cant3186 recommend this paper only to those specifically interested in seismic dataa3186 interpretation.:9L

y3187

The specific pattern recognition problem treated in this half-posthumous paperm3187 (Fu died early this year) is that of the detection of bright spots ins3187 one-dimensional seismic traces and in two-dimensional seismograms. Any3187 analyst not specialized in seismology will move very cautiously through thiss3187 strange and earthquaking land, where not only the terminology is veryg3187 peculiar, but a lot of things are supposed to be understood, without explanatio-p3187 n. What is, to begin with, a :2Obright spot”? From the conclusion of the paper,p3187 one can see that in seismograms it is possible to recognize flat spot patterns,,3187 pinchout patterns, bright spots, and so on. From the Ia definition,3187 of the bright spot can be extracted: candidate bright spot = highd3187 amplitude signals + low frequency and/or negative polarity; bright spot =-3187 candidate bright spot + continuous reflection layer. However, the exact=3187 meaning of each characteristic is rather vague, and the included seismograms do3187 not contribute to brighten the spot.

a3187

A first block diagram of a 1D syntactic pattern recognition system (why 3187 syntactic?) is presented, in order to detect candidate bright spots in testingo3187 traces selected from the input seismograms, using a tree classfication 3187 technique. A likelihood ratio test is used, unfortunately invaded by a lot ofg3187 syntactic, notational and mathematical errors (e.g., :CP:A(:Cr:A) instead off3187 :Cp:0In:A (:Cr:A) and :Cp:A:0I1 (:Cr:A); :.MC1 √:L:4Ws:9U instead of