Light field calibration and 3D shape measurement based on epipolar-space.

ABSTRACT

Light field camera calibration is much more complicated by the fact that a single point in the 3D scene appears many times in the image plane. Compared to the previous geometrical models of light field camera, which describe the relationship between 3D point in the scene and 4D light field, we proposed an epipolar-space (EPS) based geometrical model in this paper, which determines the relationship between 3D point in the scene and 3-parameter vector in the EPS. Moreover, a close-form solution for the 3D shape measurement based on the geometrical model is accomplished. Our calibration method includes an initial linear solution and nonlinear optimization with the Levenberg-Marquardt algorithm. The light field model is validated with the commercially available light field camera Lytro iIIum, and the performance of 3D shape measurement is verified by both real scene data and the data set on the internet.