A PRELIMINARY STUDY ON WAVELET-BASED CAMERA SELF-CALIBRATION

摘要

With the rapid development of digital cameras, camera calibration becomes an essential subject in photogrammetry, computer vision and optical engineering to calibrate the lens distortion of cameras. Since the 1970s, additional parameters have been widely used in self-calibration, which can determine automatically the interior orientation of a camera. Many studies focused on the use of self-calibration additional parameters in the last decades. In tradition, physical and mathematical additional parameters were proposed for analogue single-head camera calibration. However, these additional parameters might not be proper for every current diverse digital airborne camera system, such as push-broom, multi-head, virtual image composition, various image formats. In addition, many additional parameters might be highly correlated with interior orientation parameters or other parameters. Therefore, a Fourier model for self-calibration additional parameters has been developed in recent years in Stuttgart, Germany. Fourier additional parameters are orthogonal, mathematically rigorous, flexible, generic and efficient for self-calibration. The kernel functions in Fourier theory are sine and cosine functions, which are typical stationary signals. However, the lens distortion signal might be nonstationary for some modern cameras. Wavelet has not only the ability to analyze and represent nonstationary signals, but also has the same advantages as Fourier additional parameters. This paper proposes a preliminary study on establishment of a new model for camera self-calibration based on orthogonal wavelets. Moreover, this study focuses on the analysis of the prerequisites for using traditional algebraic polynomial additional parameters for camera self-calibration.