一种带预处理的RANSAC图像拼接算法

摘要

针对RANSAC算法由于迭代次数过多、精确度不高所造成的计算量增大,拼接效果不好等方面的不足.本文提出了一种带预处理的RANSAC图像拼接算法,包括图像的特征提取与粗匹配,预处理,变换矩阵的求解,图像配准和图像融合等步骤.文中重点论述了预处理的思路和方法:采用Delaunay三角剖分得到的拓扑图对数据集进行筛选,然后用欧氏距离选择RANSAC算法的初始点.通过设计拼接算法的流程和基于Matlab的实验仿真,证明该方法不但有效而且较直接RANSAC图像拼接算法减少了迭代次数、缩短了计算时间,显著提高了拼接效率.%In view of the problem that the excessive iterations lead to higher computational complexity,and the lower accuracy result in the bad effect of mosaic.In this paper,a RANSAC image stitching algorithm with preprocessing is proposed to solve above shortcomings.This algorithm includes:image feature extraction and matching,preprocessing,generating transformation matrix,image registration and image fusion.The novelty of this paper is filtering data set by topology map with Delaunay triangulation,and choosing the RANSAC algorithm initial points by the Euclidean distance.The flowchart and Matlab-based programming were designed.Comparing with the original algorithm,the experimental results show that this method has lower computational complexity and can improve the mosaic efficiently.