A Knowledge-Based Approach to Visual Information

Abstract

We propose an approach based on description logics for the representation and retrieval of visual information. We first consider objects as having shapes which are described by means of semi-algebraic sets.1 We propose a model which consists of three layers: (1) {\sf Shape Layer}, which provides the geometric shapes of image objectss (2) {\sf Object Layer}, intended to contain objects of interest and their descriptions and (3) {\sf Schema Layer}, which contains the structured abstractions of objects, i.e., a general schema about the classes of objects represented in the {\sf Object Layer}. We propose two abstract languages on the basis of description logics: one for describing knowledge of the object and schema layers, and the other, more expressive, for making queries. Queries can refer to the form dimension (i.e., information of the {\sf Shape Layer}) or to the semantic dimension (i.e., information of the {\sf Object Layer}). We show how this framework can be easily extended to accommodate the visual layer (e.g., color and texture).