（附件中含源代码）
Schoenberg (1964) showed if n≥m, the minimizer is unique and he called it natural polynomial spline.
Statisticians were interested in smoothing the spline other than interpolating  since the existence of noise.
In Practice , especially for applications in Image Registration, m=2,d=2 is pervasively used.

 

f(x,y)=d_1+d_2 x+d_3 y+∑_(i=1)^n▒〖c_i U(‖(x,y)-(x_i,y_i )‖_2 ) 〗

 

The left part ——affine part describes the behavior of f(x,y) at infinity.

The right part——locally small adjustment, bounded and asymptotically flat.

 

相关下载详见 “视觉计算研究论坛”「SIGVC BBS」：http://www.sigvc.org/bbs/thread-37-1-2.html

转载本文请联系原作者获取授权，同时请注明本文来自吴怀宇科学网博客。
链接地址：https://blog.sciencenet.cn/blog-4099-637815.html

上一篇：[CV论文读讲] Feature selection and extraction——mainly PCA
下一篇：[CV论文读讲] 3d pose tracking with linear depth and bright