Extrinsic Self-Calibration of Time-of-Flight Cameras using a Combination of 3D and Intensity Descriptors

Abstract

Time-of-Flight (ToF) cameras are able to simultaneously record intensity and depth images at a high frequency. Many applications require images that are recorded from different viewpoints. In order to consolidate the recorded data into a common coordinate system, the extrinsic calibration between the cameras needs to be known. From a practical point of view this calibration should be accomplished without any user interaction or artificial calibration objects. Classical approaches for extrinsic self-calibration fail to extract correct point correspondences and do not exploit the important information provided by the depth images. In this paper we discuss the characteristics of extrinsic ToF camera calibration and present a descriptor combination for the extraction of 3D point correspondences. Several experiments on real data demonstrate the robustness and high accuracy of our approach. Our method outperforms the state-of-the-art approach for point correspondence extraction in classical camera images.