图像拼接中相机镜头非线性畸变的校正  

Nonlinear aberration correction of lens in image mosaic

作　　者：兰海滨[1] 王平[2] 龙腾[1] 

机构地区：[1]北京理工大学信息科学技术学院,北京100081 [2]国防科技大学电子科学与工程学院,湖南长沙410073

出　　处：《光学精密工程》2009年第5期1196-1202,共7页Optics and Precision Engineering

基　　金：国家重点实验室基金资助项目(No.9140A01010507KG01)

摘　　要：为满足大尺寸图像无缝拼接的需要,提出了数字相机镜头非线性畸变的高精度校正方法。分析了现有棋盘格图像特征点检测算法的不足,提出了一种新的基于棋盘格图像结构特征的特征点检测算法。该算法巧妙地利用棋盘格图像的像素点相对于方格交点对称、相对于方格交线反对称这一特点,计算交点附近区域每一像素点的对称度和反对称度,并以此为依据来判定特征点的准确位置。将该算法与多项式变换技术以及双线性插值技术相结合,提出了数字相机图像非线性畸变的精确校正方法。实验结果表明,校正后图像的横向误差均值为0.43个像素,纵向坐标误差均值为0.36个像素。该方法精度高、计算简单,对噪声、透视、畸变等鲁棒性强,便于特征点检测的自动化处理。通过对实际地图图像的校正和拼接处理,证明了该方法的有效性。A correction method with high precision for Digital Camera（DC） lenses is proposed to meet the needs of the seamless mosaic of large scale images. The shortcomings of present algorithms are analyzed and a kind of detecting algorithm is proposed. In consideration of the structural characteristics of a chessboard pattern that the pixels in the images are central symmetrical for the points of intersection and antisymmetric for the lines of intersection of every two adjoining square panes, the precise coordinates of feature points can be decided by calculating the symmetry index and antisymmetry index for every pixel. Then, combined with the polynomial transform and bilinear interpolation techniques, a high-precision correction method for nonlinear aberration of DC images is put forward. The experiment shows the mean of horizontal error is 0.43 pixel and the mean of vertical error is 0.36 pixel after correction. These results show that proposed method has advantages in simple calculation, robust to noise, mosaic projective and distortion aberration and convenient for automatic detection. The correction and for a map photo is carried out, results prove further the effectiveness of the method by its application in the real map photo mosaic.

关 键 词：图像拼接 畸变校正 棋盘格 特征点检测 双线性插值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]