畸变图像的有效配准算法

Effective matching algorithm of distorted images

摘要:

镜头畸变是导致图像质量降低的主要因素之一.通常在图像存在畸变情况时,要先通过图像畸变校正方法来减小畸变对图像的影响,但校正过程较为复杂.本文提出了一种新的畸变图像的有效配准算法,算法分为初始匹配和松弛优化匹配两步,初始匹配算法采用圆投影匹配算法,松弛匹配算法则以向量旋转角的旋转和缩放不变性特征作为初始匹配对的优化准则,首先计算向量旋转角度差值,然后通过构造支持度函数和剔除策略选出最优匹配对.实验证明,该算法在图像非线性畸变误差较大的情况下也能够实现正确匹配.

Abstract:

The image distortion caused by lens reduces the quality of images and the processing of image correction is complex.A new effective matching algorithm of distorted images is proposed in this paper,which has two steps of the initial matching and the relaxation optimization matching.The circular projection method is applied in the initial matching step and the algorithm with characteristics of anti-rotation and anti-scale of rotating angle of vectors is applied as the optimization criterion of matching pairs.The value difference of the rotating angle of vectors should be calculated at first,and then the optimal matching pairs can be selected by using of the support function and the rules of rejection.Experimental results with a large number of images prove that the algorithm is still effective in the condition of large nonlinear distortion error of images.