A Novel Technique of Image-Based Camera Calibration in Depth-from-Defocus

Abstract

In this paper, a novel camera parameters calibration algorithm is proposed by exploiting defocus information. The proposed algorithm is based on two defocus images of the same scene obtained by changing camera's aperture numbers. Both images can be arbitrarily blurred. The blur difference between the two defocused images was estimated. Combining with imaging geometry of the thin lens, the corresponding camera parameters can be calibrated. This proposed algorithm remove the limit of calibration algorithm given by Soon-Yong Park that one of the images used in calibration must be focused. And it is valid without process of scale normalization. The calibrated parameters yield consistent results in depth estimation. Experimental results on synthetic and real images are presented to demonstrate the effectiveness of the proposed algorithm.