Computational theory of middle-level visual processing and the functional analysis

2018 Fiscal Year Final Research Report

• Project/Area Number: 16K00204
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Cognitive science
• Research Institution: The University of Electro-Communications
• Principal Investigator: Satoh Shunji 電気通信大学, 大学院情報理工学研究科, 准教授 (50333844)
• Project Period (FY): 2016-04-01 – 2019-03-31
• Keywords: 視覚モデル / MT野 / border ownership / シミュレーション / 視覚野 / 視覚心理物理実験

Outline of Final Research Achievements

Our visual system execute various information processing for retinal images, e.g. speed estimation, surface estimation, determining the depth order of occluding objects. The main purpose of the research is mathematical formation of such middle-level visual processing.
A mathematical model for speed perception is given by an improved Lucas-Kanade method. This model is quantitatively plausible because its properties are consistent with human visual properties for illusory patterns. Similar research approach was applied to the process of estimation of the angles of slant surfaces. Moreover, I found that electromagnetism is applicable for describing the visual processing for "depth-order-calculation."

Free Research Field

視覚計算論

Academic Significance and Societal Importance of the Research Achievements

中次視覚機能の定式化はそれ自体が困難である場合が多い．本研究では画像工学的観点や，これまで用いられなかったアイディア（電磁気学の基本や知見）によっていくつかの中次視覚機能を定式化し，シミュレーションによって式の妥当性を評価した．
視覚機能の定式化とシミュレーションは，知覚の予測を可能にする．例えば錯視を引きを超すパターンとひきおこさないパターンの区別が可能となる．また，運転者の視知覚を予測するための基盤的知見を与えたことにもなる．