基于Voronoi几何划分和层次化建模的纹理影像分割

doi:10.3969/j.issn.1000-436x.2014.06.011

摘要/Abstract

摘要：

将基于像素MRF分割方法拓展到基于地物目标几何约束的区域MRF分割，提出了一种基于区域和统计的纹理影像分割方法，其基本思想是利用Voronoi划分技术将影像域划分为若干子区域。在此基础上，采用二值高斯马尔科夫随机场（BGMRF, bivariate Gaussian Markov random field）模型，静态随机场模型和Potts模型从邻域、区域及全局层次描述影像的纹理结构，并将该纹理结构模型纳入贝叶斯框架；依据贝叶斯定理构建纹理影像分割模型；利用metropolis-hastings (M-H)算法进行模型参数估计，并依据最大后验概率（MAP, maximum a pos-terior）准则进行优化，从而完成纹理影像分割。为了验证所提出方法的正确性，分别对合成纹理影像，真实纹理影像及遥感影像进行了分割实验，定性和定量的测试结果验证了提出方法的有效性、可靠性和准确性。

关键词: 纹理分割, Voronoi划分, 二值高斯马尔科夫随机场, 贝叶斯定理, 最大后验概率

Abstract:

A regional and statistical based algorithm for texture image segmentation was proposed. The Voronoi tessella-tion was used for partitioning the domain of an image into sub-regions corresponding to the components of homogenous regions, to which the texture image needs to be segmented. Bivariate Gaussian Markov random field (BGMRF) model, static random field, and potts model were employed to characterize the interactions between two neighbor pixel pairs in a sub-region, and among sub-regions, respectively. Following Bayesian paradigm, a posterior distribution, which models the texture segmentation for a given texture image, was obtained. A metropolis-hastings algorithm was designed for simulating the posterior distribution. Then, texture segmentation was obtained by maximum a posterior (MAP) scheme. The proposed algorithm was tested with both of synthesized and real texture images. The results are qualitatively and quantitatively evaluated and show that the proposed algorithm works well on both of texture images.

Key words: texture segmentation, Voronoi tessellation; bivariate Gaussian Markov random field (BGMRF), Bayesian inference; maximum a posterior (MAP)

赵泉华,李玉,何晓军,宋伟东. 基于Voronoi几何划分和层次化建模的纹理影像分割[J]. 通信学报, 2014, 35(6): 82-91.

Quan-hua ZHAO,Yu LI,Xiao-jun HE,Wei-dong SONG. Voronoi tessellation and hierarchical model based texture image segmentation[J]. Journal on Communications, 2014, 35(6): 82-91.

图/表 18

图1

图2

图3

表1

图4

图5

图6

图7

图8

表2

表3

图9

图10

图11

图12

图13

图14

图15

参考文献 23

[1]	宋晓峰，王爽刘芳 . 基于区域MRF和贝叶斯置信传播的SAR图像分割[J]. 电子学报， 2010,38(12): 2810-2816. SONG X F , WANG S , LIU F . SAR image segmentation using Markov random field based on regions and Bayes belief propagation[J]. Acta Electronica Sinica, 2010,38(12): 2810-2816.
[2]	KIM T H , EOM I K , KIM Y S . Multiscale Bayesian texture segmentation using neural networks and Markov random fields[J]. Neural Computing and Applications, 2009,18(2): 141-155.
[3]	JOBANPUTRA R , CLAUSI A . Preserving boundaries for image texture segmentation using grey level co-occurring probabilities[J]. Pattern Recognition, 2006,39(2): 234-245.
	DENG H W , CLAUSI D A . Unsupervised image segmentation using a simple MRF model with a new implementation scheme[J]. Pattern Recognition, 2004,37(12): 2323-2335.
[1]	王文辉，冯前进刘磊陈等. 基于类自适应高斯-马尔可夫随机场模型和EM算法的MR图像分割[J]. 中国图象图形学报， 2008,13(3): 487-492. WANG W H , FENG Q J , LIU L , et al. Segmentation of Bra in MR images through class-adaptive Gauss-Markov random field model and the EM algorithm[J]. Journal of Image and Graphics, 2008,13(3): 487-492.
[6]	UNSER M . Local linear transforms for texture measurements[J]. Signal Processing, 1986,11(1): 61-79.
[7]	CARIOU C , CHEHDI K . Unsupervised texture segmentation / classi-fication using 2-D autoregressive modeling and the stochastic expecta-tion-maximization algorithm[J]. Pattern Recognition Letters, 2008,29(8): 905-917.
[8]	KOKKINOS I , EVANGELOPOULOS G , MARAGOS P . Texture analysis and segmentation using modulation features, generative models and weighted curve evolution[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009,31(1): 142-157.
[9]	HSU T I , KUO J L , WILSON R . A multiresolution texture gradient method for unsupervised segmentation[J]. Pattern Recognition, 2000,33(11): 1819-1833.
[10]	BOVIK A C . Analysis of multichannel narrow-band filters for image texture segmentation[J]. IEEE Transactions on Signal Processing, 1991,39(9): 2025-2043.
[11]	LI M , STAUNTON R C . Optimum Gabor filter design and local binary patterns for texture segmentation[J]. Pattern Recognition Letters, 2008,29(5): 664-672.
[12]	KIM S C , KANG T J . Texture classification and segmentation using wavelet packet frame and Gaussian mixture model[J]. Pattern Recognition, 2007,40(4): 1207-1221.
[13]	GUPTA L , MANGAI U G , DAS S . Integrating region and edge information for texture segmentation using a modified constraint satisfaction neural network[J]. Computer Vision and Image Understanding, 2008,26(8): 1106-1117.
[14]	HAMMOUCHE K , DIAF M , POSTAIRE J G . A clustering method based on multidimensional texture analysis[J]. Pattern Recognition, 2006,39(7): 1265-1277.
[15]	PARAGIOS N . A level set approach for shape-driven segmentation and tracking of the left ventricle[J]. IEEE Transactions on Medical Imaging, 2003,22(6): 773-776.
[16]	刘松涛，殷福亮 . 基于图割的图像分割方法及其新进展[J]. 自动化学报， 2012,38(6): 911-922. LIU S T , YIN F L . The basic principle and its new advances of image segmentation methods based on graph cuts[J]. Acta Automatica Sinica, 2012,38(6): 911-922.
[17]	刘茜，许刚 . 小波域隐马尔科夫树模型图像分割方法[J]. 通信学报， 2007,28(8A): 87-91. LIU Q , XU G . Image segmentation using wavelet domain hidden Mar-kov tree models[J]. Jounal on Communications, 2007,28(8A): 87-91.
[18]	齐美彬，杨立宾蒋建国 . 自适应权值的MRF分割与跟踪方法[J]. 中国图象图形学报， 2011,16(4): 572-579. QI M B , YANG L B , JIANG J G . MRF segmentation and tracking al-gorithm based on adaptive weight[J]. Journal of Image and Graphics, 2011,16(4): 572-579.
[19]	XIA G S , HE C , SUN H , et al . Integration of synthetic aperture radar image segmentation method using Markov random field on region adjacency graph[J]. IET Radar, Sonar＆Navigation, 2007,1(5): 348-353.
[20]	OKABE A , BOOTS B , SUGIHARA K , et al. Spatial Tessellations:Concepts and Applications of Voronoi Diagrams (Second Edition)[M]. Chichester: John Wiley＆Sons, 1999. 40-105.
[21]	LI S Z . Markov Random Field Modeling in Image Analysis (Third Edition)[M]. London:Springer-Verlag, 2009.
[22]	GEMAN S , GEMAN D . Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1984,6(6): 721-741.
[23]	HASTINGS W K . Monte carlo sampling methods using Markov chains and their application[J]. Biometrika, 1970,57(1): 97-109.

常量名	值
m	128
η	1.0
μ_μ	128
σ_μ	64
α_σ	0.5
β_σ	0.25
μ_ρ	2.0
σ_ρ	1.0
ε_μ	1.0
ε_θ	1/64
ε_σ	1/256
T	4 000

区域编号	μ_r/ μ_s/ μ_m	σ_r/ σ_s/ σ_m	θ₁	θ₂	θ₃	θ₄
C1	54 / 53 / 43	9.11 / 9.15 / 9.94	0.15	0.05	0.33	0.23
C2	121/120/112	10.97/11.0 /7.85	0.28	0.25	0.50	0.37
C3	134/134/ 138	5.07 / 4.91 / 4.34	0.36	0.21	0.25	0.76
C4	160/160/167	11.73/11.71/7.48	0.00	0.00	0.11	0.04
C5	209/207/225	9.74/10.15/9.17	0.03	0.09	0.07	0.15

	C1_s/ C1_m	C2_s/ C2_m	C3_s/ C3_m	C4_s/ C4_m	C5_s/ C5_m	产品精度/%
C1_s/ C1_m	3 556 / 3 241	7 / 823	34 / 2	0 / 972	2 / 2	98.81 / 63.31
C2_s/ C2_m	64 / 366	3 479 / 941	35 / 414	39 / 302	0 / 18	96.19 / 46.11
C3_s/ C3_m	0 / 16	1 / 696	1 846 / 1 523	0 / 184	0 / 65	99.95 / 61.31
C4_s/ C4_m	0 / 9	27 / 780	19 / 43	3 542 / 347	7 / 507	98.53 / 20.58
C5_s/ C5_m	13 / 1	0 / 274	51 / 3	26 / 1 802	3 636 / 3 053	97.58 / 59.48
用户精度/%	99.88 / 89.21	99.00 / 26.78	93.00 / 76.73	98.20 / 9.62	99.75 / 83.76	98.02 / 55.57

[1]	曾浩, 母王强, 杨顺平. 高机动目标跟踪ATPM-IMM算法[J]. 通信学报, 2022, 43(7): 93-101.
[2]	余盼,李斌,赵成林. 基于能量检测的异步感知算法[J]. 通信学报, 2017, 38(3): 165-173.
[3]	赵泉华，李玉，何晓军，宋伟东. 基于Voronoi几何划分和层次化建模的纹理影像分割[J]. 通信学报, 2014, 35(6): 11-91.
[4]	王利娇，张林，杨玲玲，彭立，符达伟. 具有低复杂度的LDPC已编码非相干酉空时调制系统设计[J]. 通信学报, 2014, 35(2): 19-152.
[5]	王利娇,张林,杨玲玲,彭立,符达伟. 具有低复杂度的LDPC已编码非相干酉空时调制系统设计[J]. 通信学报, 2014, 35(2): 146-152.
[6]	徐鹏飞,陈志刚,邓晓衡. 无线传感器网络中的近似Unit Delaunay功率控制算法[J]. 通信学报, 2013, 34(2): 170-176.
[7]	颜伟,蔡跃明,潘成康. 双向中继信道中物理层网络编码的检测[J]. 通信学报, 2012, 33(2): 82-86.
[8]	徐鹏飞,陈志刚,邓晓衡. 无线传感器网络中的分布式Voronoi覆盖控制算法[J]. 通信学报, 2010, 31(8): 16-25.
[9]	高荣春,韩纪庆,张磊. 说话人识别中基于最大后验概率的通道补偿方法[J]. 通信学报, 2009, 30(3): 99-103.
[10]	程翔,袁东风. 基于并行均衡框架的Turbo均衡[J]. 通信学报, 2006, 27(9): 14-20.