A novel approach to on-site camera calibration and tracking for MR pre-visualization procedure

Abstract

This paper presents camera calibration and tracking method for mixed reality based pre-visualization system for filmmaking. The proposed calibration method collects environmental information required for tracking efficiently since the rough camera path and target environment are known before actual shooting. Previous camera tracking methods using natural feature are suitable for outdoor environment. However, it takes large human cost to construct the database. Our proposed method reduces the cost of calibration process by using fiducial markers. Fiducial markers are used as reference points and feature landmark database is constructed automatically. In shooting phase, moreover, the speed and robustness of tracking are improved by using SIFT descriptor. © 2009 Springer Berlin Heidelberg.