基于迭代的非刚性点阵配准算法

doi:10.11896/j.issn.1002-137X.2016.6A.055

摘要/Abstract

摘要： 提出的非刚性点阵配准算法把一种鲁棒性全局和局部多特征用于对应关系评估,并结合高斯混合模型进行空间变换更新。首先,定义两个距离特征,分别测定两个点阵间的全局和局部几何结构差异,这两个特征形成了一种基于能量优化方程的多特征,通过最小化此多特征,可以灵活地评估点阵间的对应关系。其次,设计一种基于高斯混合模型的空间变换能量方程,同时借助L-2距离最小化方法将其最小化,以此改善空间变换更新。最后,采用轮廓配准和图像特征点配准测试了算法的性能,并与其他4种先进方法进行了对比,该算法在大部分实验中展现了最好的配准效果。

关键词: 非刚性点阵配准,多特征,对应关系评估,高斯混合模型,空间变换更新

Abstract: We proposed a non-rigid point set registration algorithm.It uses a robustly global and local multi-feature for corrspendence estimating,and combined with the Gaussian mixture model for transformation updating.Firstly,to mea-sure global and local structural diversities,we introduced two distance features,among two point sets,respectively.Then,the two features formed a multi-feature based cost matrix.It provides a flexible approach to estimate correspondences by minimizing the global or local structural diversities.Finally,we designed a Gaussian mixture model based energy function for refining the transformation updating,and it was minimized by the L2 distance minimization.By contour registration,sequence and real images,we tested the performance of the algorithm and compared against four state-of-the-art methods.This algorithm shows the best alignments in all most of the experiments.

Key words: Non-rigid point set registration,Multi-feature,Correspondence estimating,Gaussian mixture model,Transformation updating

周红玉,杨扬,张愫. 基于迭代的非刚性点阵配准算法[J]. 计算机科学, 2016, 43(Z6): 226-231. https://doi.org/10.11896/j.issn.1002-137X.2016.6A.055

ZHOU Hong-yu, YANG Yang and ZHANG Su. Non-rigid Point Set Registration Algorithm Based on Iteration[J]. Computer Science, 2016, 43(Z6): 226-231. https://doi.org/10.11896/j.issn.1002-137X.2016.6A.055

参考文献

[1] Chui H,Rangarajan A.A new algorithm for non-rigid pointmatching[J].Computer Vision and Image Understanding,2003,89:114-141
[2] Rangarajan A,Chui H,Bookstein F.The softassign procrustes matching algorithm[C]∥Information Processing in Medical Ima-ging (IPMI’97).1997:29-42
[3] Chui H,Rambo J,Duncan R,et al.Registration of cortical anatomical structures via robust 3dpoint matching[C]∥Information Processing in Medical Imaging (IPMI’99).1999:168-181
[4] Gold S,Rangarajan A,Lu C,et al.New algorithms for 2-d and 3-d point matching:pose estimation and correspondence [J].Pattern Recognition,1998,31:1019-1031
[5] Yuille A.Generalized deformable models statistical physics and matching problems[J].Neural Computation,1990, 2:1-24
[6] Bookstein,Fred L.Principal warps:thin-plate splines and the decomposition of deformations[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,1989,11(6):567-585
[7] Myronenko A,Song X,Carreira-Perpinan M.Non rigid point set registration:Coherent point drift[J].Advances in Neural Information Processing Systems,2010,32(12):1009-1016
[8] Yuille A,Grzywacz N.A mathematical analysis of the motioncoherence theory[J].International Journal of Computer Vision,1989,3:155-175
[9] Myronenko A,Song X.Point set registration:Coherent pointdrift[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32:2262-2275
[10] Greengard L,Strain J.The fast gauss transform[J].Journal of Scientific and Statistical Computing,1991,12:7994
[11] Markovsky I.Structured low-rank approximation and its applications[J].Automatica,2008,44:891-909
[12] Jian B,Vemuri B.Robust point set registration using gaussian mixture models[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2011,33:1633-1645
[13] Ruschendorf L,Rachev S.A characterization of random variables with minimum l2 distance [J].Journal of Multivariate Analysis,1990,32:48-54
[14] 周志勇,薛维琴,郑健,等.基于t分布混合模型的点集非刚性配准算法[J].光学精密工程,2013,21(9):2405-2420
[15] 周志勇,李莉华,郑健,等.含局部空间约束的t分布混合模型的点集配准[J].自动化学报,2014,40(4):683-696
[16] Ma J,Zhao J,Tian J,et al.Robust estimation of nonrigid transformation for point set registration[C]∥IEEE Conference on Computer Vision and Pattern Recognition.2013,9:2147-2154
[17] Ma J,Zhou H,Zhao J,et al.Robust Feature Matching for Remote Sensing Image Registration via Locally Linear Transforming[J].IEEE Transactions on Geoscience & Remote Sensing, 2015,53(12):6469-6481
[18] Scott D.Parametric statistical modeling by minimum integrated sqaure error[J].Technometrics,2001,43:274-285
[19] 祝继华,尹俊,邗汶锌,等.面向低维点集配准的高效最近邻搜索法[J].模式识别与人工智能,2014(12):1071-1077
[20] Wang G,Wang Z,Chen Y,et al.A robust non-rigid point set registration method based on asymmetric gaussian representation[J].Computer Vision and Image Understanding,2015,1:67-80
[21] Jonker R,Volgenant A.A shortest augmenting path algorithm for dense and sparse linear assignment problems[J].Computing,1987,8:325-340
[22] Miller M,Stone H,Cox I.Optimizing murty’s ranked assignment method[J].IEEE Transactions on Aerospace and Electronic Systems,1997,33:851-862
[23] Aronszajn N.Theory of reproducing kernels[J].Trans.of the American Mathematical Society,1950,68:337-404
[24] Micchelli C A,Pontil M.On learning vector-valued function[J].Neural Computation,2005,17:177-204
[25] Zhao J,Ma J,Zhang D.A robust method for vector field learning with application to mismatch removing[C]∥CVPR.2011,vol.32:2977-2984
[26] Rifkin G Y R,Poggio T.Regularized least-squares classification[J].In Advances in Learning Theory:Methods,Model and Applications,2003,190(1):93-104
[27] Zhou F,De la Torre F.Factorized graph matching[J].IEEEConference on Computer Vision and Pattern Recognition,2012,23(10):127-134
[28] Leordeanu M,Hebert M.Smoothing-based optimization[J].International Conference on Computer Vision and Pattern Recognition,2008,1988(4):1-8
[29] Leordeanu M,Sukthankar R,Hebert M.Unsupervised learning for graph matching[J].Internation Journal of Computer Vision,2012,96:28-45

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed