基于SURF和KLT跟踪的图像拼接算法  

Mosaic Algorithm of Image Based on SURF and KLT Track

作　　者：罗三定[1] 贺俊耀[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《计算机工程》2010年第1期215-217,共3页Computer Engineering

摘　　要：针对现有图像拼接精度不高和速度慢的问题,提出一种图像自动拼接算法。采用特征向量实现图像序列完全自动排序,把特征向量作为图像中的运动目标,利用KLT跟踪算法计算特征点的偏移量,从而得到图像之间精确的单应性变换矩阵,给出一种基于视觉特征的色彩融合方法实现图像的无缝拼接。实验证明该算法提高了匹配的精度和速度,能够实现自动排序,并具有较好的鲁棒性。Aiming at the low precision and slow speed of registration, this paper propose an algorithm for automatic mosaic of image sequence. Using SURF vector is to judge if two images are overlapped and ascertain the relationship of the overlapped images. The detected features are regarded as moving objects, located by the KLT track algorithm, and accurate homography matrix between the two images is obtained. It proposes a color fusion method to achieve seamless splicing based on visual feature. Experimental results show that the precision and speed of registration are improved and the algorithm is robust.

关 键 词：图像拼接 图像匹配 SURF算法 图像变换 

分 类 号：TN911.73[电子电信—通信与信息系统]