This paper presents an approach for estimating and then removing image radial distortion. It works onuda single image and does not require a special calibration. The approach is extremely useful in many applications,udparticularly those where human-made environments contain abundant lines. A division model is applied, in whichuda straight line in the distorted image is treated as a circular arc. Levenberg–Marquardt (LM) iterative nonlinearudleast squares method is adopted to calculate the arc’s parameters. Then “Taubin fit” is applied to obtain the initialudguess of the arc’s parameters which works as the initial input to the LM iteration. This dramatically improves theudconvergence rate in the LM process to obtain the required parameters for correcting image radial distortion.udHough entropy, as a measure, has achieved the quantitative evaluation of the estimated distortion based onudthe probability distribution in one-dimensional θ Hough space. The experimental results on both syntheticudand real images have demonstrated that the proposed method can robustly estimate and then removeudimage radial distortion with high accuracy
展开▼