基于纯旋转的相机自标定的误差分析.pdfVIP

Error Analysis of Pure Rotation-based Self-Calibration Lei Wang Sing Bing Kang Heung-Yeung Shum Guangyou Xu Computer Dept. Microsoft Research Microsoft Research China Computer Dept. Tsinghua Univ. One Microsoft Way No. 49 Zhichun Rd., Haidian Tsinghua Univ. Beijing, P.R. China Redmond, WA, USA Beijing, P.R. China Beijing, P.R. China Abstract tions can be mitigated under certain conditions. It also shows why some methods are better than others in computing in- Self-calibration using pure rotation is a well-known tech- trinsic camera parameters. In particular, we computed the nique, and has been shown to be a reliable means for recov- errorbounds for some practical special cases, and showed the ering intrinsic camera parameters. However, in practice, unexpected results that certain derivations are independent it is virtually impossible to ensure that the camera motion of certain translational components. In our analysis, we do for this type of self-calibration is a pure rotation. In this not consider degenerate motions, as was done in [8, 2, 13]. paper, we present an error analysis of recovered intrinsic The paper is organized as follows: We start by defining camera parameters due to the presence of translation. We terms and describing the problem in Section 2. Our analysis derived closed-form error expressions for a single pair of of the 1-, 2-, 3-, and 5-parameter problem is described in images with non-degenerate motion; for multiple rotations,