Surface Reflectance Model Estimation from Daylight Illuminated Image Sequences

Accurate surface reflection models derived from existing natural scenes can be used for a variety of tasks. This paper presents a machine vision approach for determining such models. We investigate the use of simplistic models of reflection and daylight illumination to determine surface reflection properties. We attempt to determine the gloss factor of both an individual surface and a multi-faceted object when illuminated by natural light with varying sun position. Experiments are performed using synthetic image sequences of surfaces illuminated by CIE standard clear and intermediate skies, viewed from a variety of camera positions. Results show that some success can be achieved using such simple illumination models. Enhancements to the proposed method are also discussed with a view to improving system performance.