Full-field speckle pattern image correlation with B-Spline deformation function
Full-field speckle pattern image correlation with B-Spline deformation function
复制标题
DOI：
10.1007/bf02410992
复制
发表时间：
2002-09
影响因子：
2.4
通讯作者：
Pengle Cheng;M. Sutton;H. Schreier;S. Mcneill
中科院分区：
文献类型：
--
作者：
Pengle Cheng;M. Sutton;H. Schreier;S. Mcneill
本文提出了一种全场散斑图像相关方法，它可以在计算机视觉系统获得的数字图像上进行图像相关处理时直接确定完整的二维变形场。在这项工作中，B样条函数被用来表示整个图像区域的对象变形场。这是基于子集的图像相关方法的改进，通过隐式地保持子集之间的位置和导数连续性约束直到指定的顺序。B样条变形函数中的控制点变量使用Levenberg-Marquardt方法迭代优化，以实现预测和实际变形图像之间的最小差异。结果表明，该方法计算效率高，准确性和鲁棒性。该方法的一般框架可以应用于求解多维向量场的一维图像相关系统。
A full-field speckle pattern image correlation method is presented that will determine directly the complete, two-dimensional deformation field during the image correlation process on digital images obtained using computer vision systems. In this work, a B-Spline function is used to represent the object deformation field throughout the entire image area. This is an improvement over subset-based image correlation methods by implicitly maintaining position and derivative continuity constraints among subsets up to a specified order. The control point variables within the B-Spline deformation function are optimized iteratively with the Levenberg-Marquardt method to achieve minimum disparity between the predicted and actual deformed images. Results have shown that the proposed method is computationally efficient, accurate and robust. The general framework of this method can be applied ton-dimensional image correlation systems that solve for multi-dimension vector fields.