出版社：The Japanese Society for Artificial Intelligence
摘要：Although there were in-depth discussions in the 1970s on the question of whether the human visual system contains 'curvature detectors' or contour detectors, which respond to the tangents of curves [Blakemore 74], they yielded no definite conclusions. Until now, the end-stopped cell model of curve detection has been the predominant one [Dobbins 87]. However, this model detects curvature with a low degree of accuracy, so a better model is required. Long ago, people discovered that the human brain is a network of numerous neurons. The hypothesis of achieving a highly accurate calculation of curvature through a network composed of biological elements (simple cells) is readily accepted. However, neither Blakemore et al. nor Dobbins et al. explain the function of simple cells in the calculation of curvature. This article illustrates the function of simple cells in calculating curvature. Moreover, in this article we attempt to construct a computational model for describing the mechanism for calculating curvatures along suggestions of Blakemore and Over. This model gives a key for answering to a question why the Helmholtz irradiation disappears when two squares are replaced by two circles.
