基于最小二乘法的鱼眼图像矫正算法

Calibration of fisheye based on least square method

使用曲面模型是矫正鱼眼图像的常规方法,一般存在桶形畸变或枕形畸变的缺点。以椭球面模型为基础,提出一种基于最小二乘法的鱼眼图像矫正算法,并采用遍历法得到全局最优参数,从而降低因拟合点的选取造成的误差影响,提高算法的拟合精度。实验结果表明,该算法在不同方向上的畸变程度均优于基于椭球面模型的鱼眼矫正算法,具有精度高、复杂度低、稳定性佳等特点,改善了矫正的效果。

Surface model can be used for calibration of fisheye. However, the shortcoming of it may be barrel dis- tortion or pincushion distortion. To overcome it, this paper takes use of least square method and traverse method based on ellipsoid model. The former one can help us to get parameters fitting the model, and random error can be reduced under the later one. Experiments have proved the precision, complexity and stability of the above calibration of fisheye compared to ellipsoid model, and the distortion in fisheye images can be calibrated satisfactorily.