Illuminant Condition Matching in Augmented Reality: A Multi-Vision, Interest Point Based Approach

For the output of an augmented reality application to appear realistic a number of issues need to be taken into consideration. The illumination correspondence between the real and virtual components should be taken into account as well as the scene level of detail and the accuracy of alignment between the two worlds. This paper focuses on matching world illumination and photometric registration methods. It introduces a new technique that aims to utilize shadow/object interest point correspondences in order to locate and virtually reproduce real-life illuminants. The technique is attractive as it makes use of natural calibration objects in the form of natural scene geometry and associated shadows. Computational complexity is kept relatively low by using an interest point based approach. Further work to be undertaken is discussed.