SIFT特征匹配的辐射畸变图像相对校正新方法

摘要

针对常规相对辐射校正方法对未配准的主从图像对校正效果欠佳的问题,在分析了辐射畸变图像的成因与特点以及相对校正的优势之后,提出了一种基于SIFT特征匹配的相对辐射校正算法.从SIFT特征的不变性出发,分别提取主从图像对的SIFT特征并进行匹配,对匹配点的灰度值进行最小二乘回归分析,得到线性变换系数,最后通过线性变换得到校正结果.实验结果表明,此方法能够有效地减小未配准主从图像对之间的辐射差异,并在复杂光照环境下的光流场计算中得到成功的应用.%In view of the issue of traditional relative radiometric correction methods not being able to handle the effects in dealing with unregistered master-slave image couples, systematic analysis of formation causes and radio-metric distortion image characteristics was carried out and the advantages of relative corrections were investigated. Therefore, a novel relative radiometric correction method was proposed based on SIFT feature matching. Depending on their feature invariance, the SIFT features of the master-slave image couple were extracted and matched first, and then the least square regression was employed to analyze the gray values of the matching points in order to obtain the linear transformation coefficients. Finally the correction result was achieved through the linear transformation of the slave image. A series of experimental results show that the radiometric differences between the unregistered master-slave image couples may be effectively reduced by using this method. In addition, this method has been successfully applied in the calculation of an optical flow field under complex light environments.