Improving Camera Pose Estimation for Indoor Marker-less Augmented Reality

Vision-based Augmented Reality (AR) techniques rely heavily on Computer Vision algorithms for most of their tasks. It is understood that these algorithms require numerous parameters to function and their values can affect their outputs. Oftentimes the results vary greatly when different parameters were used and as a result, the performance of the AR technique that utilises them varies accordingly as well. This paper aims at improving the performance of AR techniques by employing a novel algorithm to adjust the parameters automatically during runtime. More specifically, the proposed technique works on improving the camera pose estimation stage, arguably one of the most crucial stages, of such AR systems.