基于拓扑结构的图像拼接算法

摘要

In view of the disadvantages of low accuracy and large amount of calculation of Speed-Up Robust Features(SURF) algorithm because of too much matching feature points,according to the network topology theory,this paper proposes an image mosaic algorithm based on topological structure.This algorithm defines connectivity matrix and adopts it to eliminate error matching points and poor matching points.Through exclusive-OR operation of the two matrices,the judgment matrix is obtained.After several iterative operations,the final feature point set is generated.By calculating the transformation matrix,the image mosaic is finished.The experiment result shows that the mosaic algorithm based on topological structure can improve the accuracy of matching feature points more than 28.28%.For small resolution images,time-consuming remains unchanged,but for large resolution images,time consumption is significantly reduced.%针对加速稳健特征(SURF)算法匹配特征点对较多,造成拼接精度低、计算量大的缺点,结合网络拓扑学理论提出了一种剔除误匹配的SURF改进拼接算法.算法定义并采用连通矩阵得到特征点的拓扑结构,将两幅图像的连通矩阵做异或运算得到判断矩阵,多次迭代剔除特征点集合中错误匹配点和匹配度较低的点,得到拓扑结构完全相同的特征点集合,有效减少了匹配特征点对的数量,提高了特征点对的匹配正确率.根据最终的特征点集合计算变换矩阵完成两幅图像的拼接处理.对比实验结果表明,针对同一组图像进行拼接处理,所提算法的特征点匹配正确率较SURF算法提高了28.28%以上;对于小分辨率图像拼接,耗时基本保持一致;对于大分辨率图像拼接,所提算法耗时大大减少.