相互抑制型ニューラルネットによる幾何学錯視の実現及び視空間の幾何学に関する研究

概要

論文の詳細を見る
Illusions in human vision, especially geometric illusions, have been a theme for century-lasting efforts in order to explore deep and mysterious mechanism of visual information processing in human brain. Most researches on the subject have been to explain geometric illusion phenomena under various ad-hoc psychological hypotheses for particular phenomenon and no unifying model is known. Here we address the lateral inhibition model which is the only one supported by physiological facts. This model, even well-known for quite long time, has not been accepted since no convincing evident to show that illusion can be produced without other forces. It also met difficulty such as the paradox of displacement. Another related questions still unsettled is about where the illusion occurs, is it in the retina or in the cortex? Psychologists preffered and tried hard to prove the second answer since it introduce some interesting mental process. By the metal lateral inhibition between direction detection neurons on cortex, one can explain the metric illusion such as Muller-Lyer illusion, or the angular illusion such as Zollner, Poggendorff illusions, but little can be done about the other illusions. In this paper, we first show that we can prove that the geometric illusions could occur within the retina without involving higher hierarchy of visual perception. This is done by realization of various illusion phenomena with a unifying neural network model of lateral inhibition. Then, we will applied this model for quantitative analysis of the intrinsic geometric properties such as twist or bending of the visual space in geometric illusions.
社団法人電子情報通信学会の論文
1994-09-26

相互抑制型ニューラルネットによる幾何学錯視の実現及び視空間の幾何学に関する研究

スポンサーリンク

概要

著者

関連論文

スポンサーリンク