Marker-less tracking for multi-layer authoring in AR books

Abstract

An Augmented Reality (AR) book is an application that applies AR technologies to physical books for providing a new experience to users. In this paper, we propose a new marker-less tracking method for the AR book. The main goal of the tracker is not only to recognize many pages, but also to compute 6 DOF camera pose. As a result, we can augment different virtual contents according to the corresponding page. For this purpose, we use a multi-core programming approach that separates the page recognition module from the tracking module. In the page recognition module, highly distinctive Scale Invariant Features Transform (SIFT) features are used. In the tracking module, a coarse-to-fine approach is exploited for fast frame-to-frame matching. Our tracker provides more than 30 frames per second. In addition to the tracker, we explain multi-layer based data structure for maintaining the AR book. A GUI-based authoring tool is also shown to validate feasibility of the tracker and data structures. The proposed algorithm would be helpful to create various AR applications that require multiple planes tracking. © 2009 Springer.