基于SIFT特征检测的图像拼接优化算法研究*

针对复杂场景下图像拼接,误匹配点比例较大时,传统匹配优化算法效率低,合成图像易产生鬼影等问题,在SIFT算法基础上,采用一种新的聚类方法预筛选特征点对,再用RANSAC算法精确提纯,减少算法迭代次数;并提出了改进的基于特征点的最佳缝合线与多分辨率样条法相结合的融合方法,提升了融合图像质量。实验结果表明,经过对以上两部分的改进,算法效率有较大提高,并能有效去除鬼影现象。

Research on image mosaic optimization algorithm based on SIFT feature detection

In complex scenes, the traditional image match optimization algorithm was low efficiency, and motion ghosting was often contained in the merged image. In order to solve these problems, on the basis of the SIFT algorithm, this paper adopted a new method of clustering to pre-screening feature points, then used RANSAC algorithm for purifying feature points accurately. About image fusion, based on feature point, it proposed an improved methods which combined the best suture line and multi-resolution method. Experiments prove that, the improved algorithm which is efficiency, can effectively remove the ghosting.