Outdoor Illumination Estimation in Image Sequences for Augmented Reality

The paper presents a technique for estimating the radiance of the sky and sun for outdoor, daylight illumination conditions. Shadows cast by dynamic objects are detected using color imagery, combined with depth information from a commercial stereo camera setup. Color information extracted from the detected shadows is used to estimate the radiance of the sun. The technique does not require special purpose objects in the scene, nor does it require High Dynamic Range imagery. Results are demonstrated by rendering augmented objects into real images with shading and shadows which are consistent with the real scene