基于动态阈值和全局信息的SIFT量子图像拼接

Quantum image stitching based on SIFT algorithm with dynamic threshold and global information

针对量子图像拼接使用传统尺度不变特征变换(SIFT)算法易造成特征点数量不合理和误匹配率高的问题,提岀基于动态阈值和全局信息的SIFT量子图像拼接算法.针对特征点数量问题,利用特征点数量与量子点或量子环数量呈正比例关系这一特性,通过量子点或环的密度对SIFT算法的对比度阈值进行设置,以得到合适的特征点数量.针对误匹配率问题,通过构建全局信息描述子,并与SIFT局部描述子相结合,以降低误匹配率.实验结果表明:改进的算法有效地完成了量子环、量子线和量子点图像的拼接;将量子图像的特征点有效地控制在一个合理的范围内,并将误匹配率从17.34%〜33.02%降低至10.84%〜20%,使量子图像的拼接具有更好的可靠性.

When it is applied to quantum image stitching,the traditional scale-invariant feature transform(SIFT)algorithm will result in the problems such as unreasonable feature point distribution and high mismatch rate.So an improved SIFT quantum image stitching algorithm based on dynamic threshold and global information is proposed.In the aspect of unreasonable feature point number,as the number is Jproportional to the number of quantum dots or quantum rings,the contras11hreshold of SIFT algorithm is set by calculating the density of quantum dots or rings so as to get the appropriate feature point number.In the aspec t of reducing high false mismatch rate,a global information descrip tor is cons tructed and combined with the local SIFT descrip tor.The experimental results show that the improved algorithm can effectively comple te the st itching of quantum rings,quantum wire and quantum dot images,effectively control the feature points of quantum images within a reasonable range,and reduce the mismatch rate from 17.34%~33.02% to 10.84%~20%,which makes the quantum image stitching have better reliability.