Efficient Geometric, Photometric, and Temporal Calibration of an Array of Unsynchronized Video Cameras

Abstract

Camera-arrays have become popular in many computer vision and computer graphics applications. Among all preprocessing steps, an efficient method to calibrate a large number of cameras is very much desired. The required calibration includes both the geometric and photometric calibration, which are the most common and also well studied for single camera. However, few existing efforts are devoted to camera arrays or to integrate both methods in a fully automatic way. Additionally, most existing camera array systems require or assume implicitly that all the cameras in the array are hardware-synchronized to simplify subsequent application-specific processing such as the calibration of all the cameras. While this constraint is useful, it greatly restricts the use of heterogeneous types of cameras and the configuration of cameras that could be used. In this paper, we propose a novel integrated and fully automatic solution for performing geometric, photometric and temporal calibration (synchronization) of an array of unsynchronized video cameras. In particular, our new method is based on the classic plane based calibration approach. By using a redesigned calibration pattern, the geometric, photometric and temporal calibrations are done in an integrated and extensible framework automatically. Extensive experimental results show that the new method is very easy to use and can achieve high accuracy in the calibrated parameters.