改进的RANSAC算法在图像配准中的应用

Application of improved RANSAC algorithm to image registration

为了提高图像配准的速度,提出了一种基于改进的随机抽样一致性(RANSAC)算法的快速图像配准方法。该方法首先采用Harris角点检测算法提取出参考图像和目标图像的特征角点,然后利用灰度相关性进行特征角点的匹配,最后采用基于预检测的RANSAC算法快速而精确地估计变换矩阵,进行图像配准。该算法中采用预检测的方法快速抛弃那些不是候选模型的临时模型,提高了算法的速度。同时使用随机块选取法选择样本,很好地消除外点的影响进而保证精度。实验结果表明,此方法在得到较高的精度和鲁棒性的情况下,还大幅度减少了运算量,提高了图像配准的速度。

In order to improve the speed of image registration,a fast method based on improved RANdom SAmple Consensus（RANSAC） algorithm was proposed.At first,Harris corner detector was used to extract the feature points in the reference and target images.Next,based on the proximity and similarity of their intensity,the feature points were matched.Finally,the improved RANSAC algorithm was used to estimate transform matrix more fast and accurately.To improve the speed of computation,a method of pre-detection was used to discard those temporary models that were not preview models.To delete outliers a random block selecting method was used to select samples,which improves the precisions.The experiment shows that this algorithm reduces the amount of computation largely and improves the speed of image registration when the precisions do not have notable change.