实际光学镜头所成的图象难免会有几何变形，多项式坐标变换法是进行几何修正的有效方法，但是当次数较高的时候，运算量太大，难以应用到实时图象处理系统，为此在分析多项式坐标变换算法的基础上，提出了一次多项式非均匀分片逼近算法。该算法首先将图象非均匀划分成矩形区域，在每个矩形区域内部用一个一次多项式逼近高次多项式。基于对图象畸变的主要因素即径向畸变的分析，该算法的图象划分规则能在保证逼近精度的前提下占用最少的保存模型参数的空间。该算法大大降低了运算量，将运算时间减少了近2／3，同时能很好地保证逼近精度，空间代价也限制在很小的范围内，试验结果表明该算法是图象几何修正的有效方法，具有良好的工程应用价值。

Geometrical distortion in optical lens imaging is inevitable, and the polynomial coordinate transform algorithm is useful in rectification of image geometrical distortion, but it takes too much computational time to be applied in real time image processing systems when high order polynomials are employed on condition that complex nonlinear distortions are encountered. To decreases the computational complexity, an approximation algorithm is proposed in this paper, in which the image is firstly divided into rectangles, and then an appropriate first-order polynomial is used to approximate the high order polynomial within each rectangle. Based on the analysis of the radial distortion, which is considered to be the main kind of image distortion, a novel image partition method is presented to ensure minimal memory expense for model parameters. With the approximation algorithm two-third of the computational time is saved and the memory expense is also limited within a reasonable range. Experimental results indicate that the algorithm proposed in this paper is effective and valuable in the engineering applications.