A Feature Extraction Algorithm Based on Local Picture Computation
Yoshio Hayashi, Yasuzo Suto · IEEE Transactions on Systems Man and Cybernetics · 1977
A binary picture feature extraction algorithm using picture primitive notation and Perceptrons is presented. The first step is to define picture basic segments as geometric factors. This algorithm proves that the pictorial structure can be described by the geometric factors so that such effective feature values as picture size, perimeter, picture shape, and texture can be extracted on the basis of local picture computation. The local property of picture computation is evaluated by the concept called "order" of the Perceptrons. Further, based on this feature extraction algor-ithm, an efficient picture recognition system is realized. The effectiveness and practicability of this algorithm are clarified by computer simulation.