The Interpretation of Three-Dimensional Structure from Image Curves

Abstract

General constraints on the interpretation of image curves are described and implmnented. We illustrate the use of these constraints to int erpret three-dimensional st ructure from an image up to the volumetric level. These constraints do not use any knowledge regarding the specific objects in the image, but rely on general assumptions regarding illumination, occlusion, object geometry, anrl the imaging process. They are based on coincidence assumptions- that various coincidental /eatures or alignments in an image are unlikely to arise without reason . The strength of these coincidence assumptions depends on the accuracy of the low-level description of an image. Since any one feature can be the result of pure coincidence or erron in the detection process, it is necessary to use a reasoning system which can use these hypotheses to derive consistent global interpretations, while maintaining the ability to remove the implications of hypotheses which are disproved in the face of further evidence. We demonstrate the successful interpretation of some hand-derived image curves up to the volumetric level, including the construction of a three-space model.