A New Geometrical Interpretation of Trilinear Constraints

Abstract

We give a new geometrical interpretation of the well-known algebraic trilinear constraints used in motion analysis from three views observation. We show that those algebraic equations correspond to depth errors appropriately weighted by a function of the relative reliability of the corresponding measurements. Therefore directly minimizing the algebraic trilinear equations, in the least squares sense, is a strategy that works well for estimating motion. In addition, we propose a new scheme for recovering the scale factor for motion estimation that is very insensitive to input noise. All our theoretical statements are supported by experimental results.