基于特征点的全自动无缝图像拼接方法  

Automatic seamless image mosaic method based on feature points

作　　者：李寒[1] 牛纪桢[1] 郭禾[1] 

机构地区：[1]大连理工大学计算机科学与工程系,辽宁大连116023

出　　处：《计算机工程与设计》2007年第9期2083-2085,共3页Computer Engineering and Design

摘　　要：提出了一种基于特征点的全自动无缝图像拼接方法。该方法采用对于尺度具有鲁棒性的SIFT算法进行特征点的提取与匹配,并通过引导互匹配及投票过滤的方法提高特征点的匹配精确度,使用稳健的RANSAC算法求出图像间变换矩阵H的初值并使用LM非线性迭代算法精炼H,最终使用加权平滑算法完成了图像的无缝拼接。整个处理过程完全自动地实现了对一组图像的无缝拼接,克服了传统图像拼接方法在尺度和光照变化条件下的局限性。实验结果验证了方法的有效性。An automatic seamless image mosaic method based on feature points is proposed, First a scale-invariant feature extracting algorithm SIFT is used for feature extraction and matching, In order to improve the accuracy of matching, guided complementary matching and voting filter is used. Then, the transforming matrix H is computed with RANSAC algorithm and LM algorithm. And finally image mosaic is completed with smoothing algorithm. The method implements automatically and avoids the disadvantages of traditional image mosaic method under different scale and illumination conditions. Experimental results show that the image mosaic method is stable and effective.

关 键 词：图像拼接 SIFT特征点 引导互匹配 随机抽样一致算法 变换矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]