Abstract
Current augmented and mixed reality systems suffer a lack of correct illumination modeling where the virtual objects render the same lighting condition as the real environment. While we are experiencing astonishing results from the entertainment industry in multiple media forms, the procedure is mostly accomplished offline. The illumination information extracted from the physical scene is used to interactively render the virtual objects which results in a more realistic output in real-time. In this paper, we present a method that detects the physical illumination with dynamic scene, then uses the extracted illumination to render the virtual objects added to the scene. The method has three steps that are assumed to be working concurrently in real-time. The first is the estimation of the direct illumination (incident light) from the physical scene using computer vision techniques through a 360 degrees live-feed camera connected to AR device. The second is the simulation of indirect illumination (reflected light) from the real-world surfaces to virtual objects rendering using region capture of 2D texture from the AR camera view. The third is defining the virtual objects with proper lighting and shadowing characteristics using shader language through multiple passes. Finally, we tested our work with multiple lighting conditions to evaluate the accuracy of results based on the shadow falling from the virtual objects which should be consistent with the shadow falling from the real objects with a reduced performance cost.