Automatic camera self-calibration for immersive navigation of free viewpoint sports video

[Paper] Sports Camera Calibration using Flexible Intersection Selection and Refinement

“…After that, horizontal and vertical lines are selected to identify intersections. 16) When the line detection fails, a homography matrix is not calculated due to the lack of the number of intersections.…”

[Paper] Quality Improvement for Real-time Free Viewpoint Video Using View-dependent Shape Refinement

“…To reduce the calibration errors, the methods 28)29) focus on improving the accuracy of the intersection detection. Yao et al 28) classify detected field lines into horizontal and vertical directions to find intersections. In addition, the flexible intersection selection method 29) improves the accuracy of the field line-based method 28) .…”

“…Yao et al 28) classify detected field lines into horizontal and vertical directions to find intersections. In addition, the flexible intersection selection method 29) improves the accuracy of the field line-based method 28) . However, these methods require many field lines to estimate accurate camera parameters.…”

“…, f K−1 (u, v)} is manually clicked on a screen for all cameras, where K indicates the number of all visible feature points in the image. After obtaining the set of feature points F c by a human operator, internal and external camera parameters are estimated by using the conventional camera calibration method 28) using field line intersections. A set of the feature points in 3D stadium coordinates G c = {g 0 (x, y, z), g 1 (x, y, z), .…”

“…As described in subsection 2. 2, the method 28) detects an intersection automatically. However, parameter tuning is required, and the applicable scene is limited.…”