Estimating Intrinsic Parameters of Cameras using Two Arbitrary Rectangles

In this paper, we propose new camera calibration methods assuming a static camera. Two corresponding imaged rectangles whose aspect ratios are unknown are sufficient to calibrate a camera. By warping the images properly, we show that the information from the imaged rectangles can be transformed to the form of camera constraints. Based on this results, we propose two methods, one for three or more images and the other for only two images. The proposed methods are verified with synthetic and real images, and the results are comparable with less assumptions on cameras and on scenes.