Weighted hough transform for directional light sources estimation from multi-view images

Abstract

An improved method for light sources estimation from shading is proposed in this paper. This method indicates the advantage of multi-view images rather than single view image. Even though there are several methods that can be used to estimate light sources from single view image, data from single view may be not accurate for all light source estimations. The shading on a Lambertian surface sphere is used to detection critical paths and the directions of directional light sources can be estimated from these paths. Hough Transform is used to find critical paths from critical points in images. Rather than using the standard Hough Transform, we improve vote count processes by determining a weight value for each critical point in Hough space. Weight values are calculated from angles between surface normal at each critical point and view vector from cameras. The experimental results demonstrate that the proposed method with information obtained from multi-view images yields better performance than the method with a single-view image.