基于L-M优化算法的多幅岩心荧光图像拼接

在多幅岩心荧光图像的拼接中,为了减小累计误差造成的拼接畸变,提出了一种基于L-M(Levenberg-Mar-quardt)算法的全局优化方法。该方法是在每一幅图像的初始球面投影参数的基础上,采用L-M算法进行迭代计算,在迭代过程中根据增长率δ的取值实时调整阻尼因子μ,从而不断更新迭代方向和迭代步长,求出最优投影参数集,使所有图像的全部的特征点经球面变换后的误差距离总和最小,实现全局配准。把该算法应用到多幅岩心荧光图像的拼接中,解决了累计误差造成的畸变问题,获得了满意的拼接效果。

Multiple Core Fluorescence Images Stitching Based on L-M Optimization Algorithm

Considering the distortion caused by the accumulative error in multiple core fluorescence image stitching,a global optimization method was proposed based on L-M(Levenberg-Marquardt) algorithm.In order to achieve global registration,which can minimize the sum of error distance for all feature point after spherical transforming in all image,based on the initial parameters of the spherical projection for each image,L-M algorithm was used to do iterative calculation.During the iterative calculation,the damping parameter μ was real-time adjusted due to the gain ratio δ so that the iteration direction and the size of step were continuously updated.Then the optimal set of projection parameter was obtained finally.The algorithm was applied to the multiple core fluorescence images stitching,solved the problem of the distortion of stitched image caused by the accumulative error and obtain satisfactory results.