Reuter,A., Seidel,H.-P., Ihrke,I.
Abstract:
Current research is targeting the estimation and correction of lens imperfections. Due to their spatial variation, these are often encoded as a set of point spread functions, distributed over the image domain. To perform the calibration process, checkerboard patterns are typically employed. However, previous work does not exploit all benefits of using a checkerboard. In our approach, we use a checkerboard to estimate the amount of out-of-focus blur to make the calibration process more robust to unfocussed patterns. Due to this, we are able to calibrate the lens aberrations even if the calibration pattern is out-of-focus. Further, we propose a blur-stable checkerboard tracking approach to improve the robustness of the detection process. Our detection approach is fast enough to run in realtime on videos.