Lens calibration without pre-designated distortion model and corresponding points

A new lens calibration method which is suitable for all kinds of cameras is presented. Based on the linear feature in the image, this method does not require the determination of point correspondence between a scene and the image of the scene. By applying B-spline surface fitting, all kinds of lens distortion models can be simulated. The control points of B-spline surface are estimated iteratively to extract linear feature from the distorted image. Compared with traditional calibration method, which requires either corresponding point pairs or distortion model of lens, or both of them, this method has its unique advantage. Experimental results show the effectiveness of the method. Foundation Item: Supported by the Chen Guang Project Biography: CHEN Shu-zhen (1946), female, Professor, Research direction digital image processing, computer vision.