Theory of Vision
Year: 2026
Type: preprint
Abstract: This paper proposes a mathematical framework called the Theory of Vision, designed to investigate the limitations of visual observation and the conditions under which information about an object's position, depth, and identity can be recovered. The theory begins with the fundamental observation that a single observer, relying solely on angular information, cannot generally determine both the distance and physical extent of an object. This underdetermination serves as the foundation for a broader framework addressing visual ambiguity and information recovery. The framework is developed through several interconnected components: Single-Observer Geometry: Establishes the fundamental limitations of depth and size determination from a single observation point. Two-Observer Disparity: Introduces an asymmetric two-observer configuration and derives equations for recovering an object's depth and transverse position. Degenerate Configurations: Identifies conditions under which depth recovery becomes impossible, including coincident observation points and collinear configurations. Variable Fields of View: Formalizes observation regions through independently adjustable positions, orientations, and angular apertures. Reflective Geometry: Extends the framework to mirrors using virtual points and geometric reflection. Angular Resolution: Introduces the Window Spider Theorem, which examines the limits of distinguishing an object's position across a transparent surface. Transmitted Luminance: Develops an independent observational channel based on light attenuation through partially transmissive media, leading to the Paper Spider Theorem. Occlusion and Visibility: Introduces the Treasure Hunting Theorem, distinguishing between the inability to detect an occluded object and the inability to identify an object through a semi-transparent obstruction. The framework is organized around two primary observational channels: angular disparity and transmitted luminance. These channels describe different mechanisms through which visual information can be obtained or become insufficient. The paper also discusses potential extensions involving tactile observation, multiple observers, and wave-optical phenomena. The primary objective of this work is to establish a unified geometric and radiometric perspective on visual observation, focusing on the relationship between observable information, ambiguity, and the conditions required for recovering information about physical objects.
Author: Matsuoka, Ren
Language: English
Topic: Advanced Vision and Imaging
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235723352
