Estimating vanishing point from the angular correlation of spectral texture distribution

This paper describes a novel method for recovering the perspective geometry of textured surfaces from local spectral analysis. We recover 3-D orientation of planar surfaces based on angular properties of the local spectral description of the surface markings. We also show a geometrical relationship between local spectral components and the vanishing point of the texture plane. Our method is a two-step process. We commence by identifying contours which connect the local spectral moments of equal power. The direction can be estimated from the energy gradient described in the perpendicular direction to these contours. The second step is to correlate points with the same angular spectral distribution. Points with the same angular spectral configuration will lie on a line pointing to a vanishing point. This allows us to estimate the direction of the vanishing point at different positions on the image plane. In this way we can use a sample of spectral moments to estimate the vanishing point location.