摘要
针对基于图像特征点匹配的图像拼接存在的缺点,文中提出了一种自动图像拼接新算法。该算法使用了改进的RANSAC算法来消除局外点对全局变换参数估计的影响,而在初始参数估计问题上,则采用分层渐进匹配的机制及梅林变换,从而有效地解决了方位变动较大的图像带来大的初始参数估计的困难,同时有效地提高了计算效率。实验表明,使用该技术后图像拼接效果较理想,鲁棒性好,真实感强,实际应用价值较高。
By analyzing the disadvantages of the keypoint - based image stitching, a new automatic image - stitching algorithm is pro-osed. An improved RANSAC algorithm is used for getting rid of influence of the bad points. In the problem of the parameter estimation, a layered progressive matching scheme and MeUion transformation are employed in the initial parameter estimation when orientation of the images is great. The above scheme also improved computing efficiency. Experimental image-stitching results are better satisfactory, better robustness and stronger sense of reality. It is high valuable in practice.
出处
《计算机技术与发展》
2009年第6期36-38,42,共4页
Computer Technology and Development
基金
重庆大学研究生科技创新基金(200801A1B0290288)
关键词
特征点
图像拼接
鲁棒估计
梅林变换
keypoint
image - stitching
robust estimation
Mellion transformation