A Self-calibration Algorithm Based on a Unified Framework for Constraints on Multiple Views

In this paper, we propose a new self-calibration algorithm for upgrading projective space to Euclidean space. The proposed method aims to combine the most commonly used metric constraints, including zero skew and unit aspect-ratio by formulating each constraint as a cost function within a unified framework. Additional constraints, e.g., constant principal points, can also be formulated in the same framework. The cost function is very flexible and can be composed of different constraints on different views. The upgrade process is then stated as a minimization problem which may be solved by minimizing an upper bound of the cost function. This proposed method is non-iterative. Experimental results on synthetic data and real data are presented to show the performance of the proposed method and accuracy of the reconstructed scene.