# American Institute of Mathematical Sciences

June  2019, 13(3): 653-677. doi: 10.3934/ipi.2019030

## A dual EM algorithm for TV regularized Gaussian mixture model in image segmentation

 1 Laboratory of Mathematics and Complex Systems (Ministry of Education of China), School of Mathematical Sciences, Beijing Normal University, Beijing 100875, China 2 Department of Mathematics, Hong Kong Baptist University, Kowloon Tong, Hong Kong, China 3 Depart of Mathematics, University of Bergen, P. O. Box 7800, N-5020, Bergen, Norway

* Corresponding author: Jun Liu

Received  May 2018 Revised  October 2018 Published  March 2019

A dual expectation-maximization (EM) algorithm for total variation (TV) regularized Gaussian mixture model (GMM) is proposed in this paper. The algorithm is built upon the EM algorithm with TV regularization (EM-TV) model which combines the statistical and variational methods together for image segmentation. Inspired by the projection algorithm proposed by Chambolle, we give a dual algorithm for the EM-TV model. The related dual problem is smooth and can be easily solved by a projection gradient method, which is stable and fast. Given the parameters of GMM, the proposed algorithm can be seen as a forward-backward splitting method which converges. This method can be easily extended to many other applications. Numerical results show that our algorithm can provide high quality segmentation results with fast computation speed. Compared with the well-known statistics based methods such as hidden Markov random field with EM method (HMRF-EM), the proposed algorithm has a better performance. The proposed method could also be applied to MRI segmentation such as SPM8 software and improve the segmentation results.

Citation: Shi Yan, Jun Liu, Haiyang Huang, Xue-Cheng Tai. A dual EM algorithm for TV regularized Gaussian mixture model in image segmentation. Inverse Problems & Imaging, 2019, 13 (3) : 653-677. doi: 10.3934/ipi.2019030
##### References:

show all references

##### References:
Comparison of Dual EM-TV, Original EM-TV [26] and HMRF-EM methods for 2 classes segmentation
Enlarged red square regions in Figure 1
Comparison of Dual EM-TV, Original EM-TV [26] and HMRF-EM methods for 4 classes segmentation
Enlarged red square regions in Figure 3
An image with two regions, which are the same mean but different variance, and the histogram
Comparison of segmentation results of GMM-EM, Dual EM-TV and CV model on region with same mean but different variances image
Comparison between Dual EM-TV and HMRF-EM method
Segmentation results by Dual EM-TV in color images for two-region images
Segmentation results by Dual EM-TV in color images for multi-region images
Segmentation comparison under different noise levels. Noise standard deviation: A: 120/255, B: 240/255, C: 300/255, D: 390/255
Segmentation comparison of the same multi-region image by different noise levels. Noise standard deviation: A: 50/255, B: 60/255, C: 70/255, D: 80/255
Results of SPM8 with one slice of MR Images. First row: left image, MRI with 0% RF and 0% noise; right: MRI with 40% RF and 9% noise). Second and third rows: the first and third columns are the newsegment method for left and right images, respectively; the second and fourth columns are the proposed method for left and right images, respectively
Comparison of 2 classes segmentation for Dual EM-TV, Original EM-TV [26] and HMRF-EM method (Heart image with size $615\times 615$)
 Methods CPU time (seconds) Accuracy (SAI) Dual EM-TV 1.233 99.70% Original EM-TV [26] 9.482 98.02% HMRF-EM 4n 4.052 91.84% HMRF-EM 8n 3.052 97.38% %HMRF-EM 12n 2.653 98.69% HMRF-EM 20n 1.528 99.24%
 Methods CPU time (seconds) Accuracy (SAI) Dual EM-TV 1.233 99.70% Original EM-TV [26] 9.482 98.02% HMRF-EM 4n 4.052 91.84% HMRF-EM 8n 3.052 97.38% %HMRF-EM 12n 2.653 98.69% HMRF-EM 20n 1.528 99.24%
Comparison of 4 classes segmentation for Dual EM-TV, Original EM-TV [26] and HMRF-EM methods (4-color image with size $600\times 600$)
 Methods CPU time (seconds) Accuracy (SAI) Dual EM-TV 1.699 99.90% Original EM-TV [26] 18.502 99.02% HMRF-EM 4n 3.480 97.39% HMRF-EM 8n 3.202 99.81% %HMRF-EM 12n 2.696 99.87% HMRF-EM 20n 2.388 99.83%
 Methods CPU time (seconds) Accuracy (SAI) Dual EM-TV 1.699 99.90% Original EM-TV [26] 18.502 99.02% HMRF-EM 4n 3.480 97.39% HMRF-EM 8n 3.202 99.81% %HMRF-EM 12n 2.696 99.87% HMRF-EM 20n 2.388 99.83%
Comparison of DM in SPM8 on brain images
 noise level 5% 5% 7% 7% 9% 9% brain part WM GM WM GM WM GM Dual EM-TV 0.9380 0.9122 0.9218 0.8970 0.9035 0.8818 New Segment 0.9317 0.9099 0.9035 0.8822 0.8759 0.8536
 noise level 5% 5% 7% 7% 9% 9% brain part WM GM WM GM WM GM Dual EM-TV 0.9380 0.9122 0.9218 0.8970 0.9035 0.8818 New Segment 0.9317 0.9099 0.9035 0.8822 0.8759 0.8536
 [1] Baoli Shi, Zhi-Feng Pang, Jing Xu. Image segmentation based on the hybrid total variation model and the K-means clustering strategy. Inverse Problems & Imaging, 2016, 10 (3) : 807-828. doi: 10.3934/ipi.2016022 [2] Xiaojing Ye, Haomin Zhou. Fast total variation wavelet inpainting via approximated primal-dual hybrid gradient algorithm. Inverse Problems & Imaging, 2013, 7 (3) : 1031-1050. doi: 10.3934/ipi.2013.7.1031 [3] Juan C. Moreno, V. B. Surya Prasath, João C. Neves. Color image processing by vectorial total variation with gradient channels coupling. Inverse Problems & Imaging, 2016, 10 (2) : 461-497. doi: 10.3934/ipi.2016008 [4] Ming Yan, Alex A. T. Bui, Jason Cong, Luminita A. Vese. General convergent expectation maximization (EM)-type algorithms for image reconstruction. Inverse Problems & Imaging, 2013, 7 (3) : 1007-1029. doi: 10.3934/ipi.2013.7.1007 [5] Zhengmeng Jin, Chen Zhou, Michael K. Ng. A coupled total variation model with curvature driven for image colorization. Inverse Problems & Imaging, 2016, 10 (4) : 1037-1055. doi: 10.3934/ipi.2016031 [6] Xavier Bresson, Tony F. Chan. Fast dual minimization of the vectorial total variation norm and applications to color image processing. Inverse Problems & Imaging, 2008, 2 (4) : 455-484. doi: 10.3934/ipi.2008.2.455 [7] Ke Chen, Yiqiu Dong, Michael Hintermüller. A nonlinear multigrid solver with line Gauss-Seidel-semismooth-Newton smoother for the Fenchel pre-dual in total variation based image restoration. Inverse Problems & Imaging, 2011, 5 (2) : 323-339. doi: 10.3934/ipi.2011.5.323 [8] Yunhai Xiao, Junfeng Yang, Xiaoming Yuan. Alternating algorithms for total variation image reconstruction from random projections. Inverse Problems & Imaging, 2012, 6 (3) : 547-563. doi: 10.3934/ipi.2012.6.547 [9] Jie Huang, Xiaoping Yang, Yunmei Chen. A fast algorithm for global minimization of maximum likelihood based on ultrasound image segmentation. Inverse Problems & Imaging, 2011, 5 (3) : 645-657. doi: 10.3934/ipi.2011.5.645 [10] Jianping Zhang, Ke Chen, Bo Yu, Derek A. Gould. A local information based variational model for selective image segmentation. Inverse Problems & Imaging, 2014, 8 (1) : 293-320. doi: 10.3934/ipi.2014.8.293 [11] Sören Bartels, Marijo Milicevic. Iterative finite element solution of a constrained total variation regularized model problem. Discrete & Continuous Dynamical Systems - S, 2017, 10 (6) : 1207-1232. doi: 10.3934/dcdss.2017066 [12] Liyan Ma, Lionel Moisan, Jian Yu, Tieyong Zeng. A stable method solving the total variation dictionary model with $L^\infty$ constraints. Inverse Problems & Imaging, 2014, 8 (2) : 507-535. doi: 10.3934/ipi.2014.8.507 [13] Xiaoqun Zhang, Tony F. Chan. Wavelet inpainting by nonlocal total variation. Inverse Problems & Imaging, 2010, 4 (1) : 191-210. doi: 10.3934/ipi.2010.4.191 [14] Dominique Zosso, Jing An, James Stevick, Nicholas Takaki, Morgan Weiss, Liane S. Slaughter, Huan H. Cao, Paul S. Weiss, Andrea L. Bertozzi. Image segmentation with dynamic artifacts detection and bias correction. Inverse Problems & Imaging, 2017, 11 (3) : 577-600. doi: 10.3934/ipi.2017027 [15] Matthew S. Keegan, Berta Sandberg, Tony F. Chan. A multiphase logic framework for multichannel image segmentation. Inverse Problems & Imaging, 2012, 6 (1) : 95-110. doi: 10.3934/ipi.2012.6.95 [16] Ye Yuan, Yan Ren, Xiaodong Liu, Jing Wang. Approach to image segmentation based on interval neutrosophic set. Numerical Algebra, Control & Optimization, 2019, 0 (0) : 0-0. doi: 10.3934/naco.2019028 [17] Rinaldo M. Colombo, Francesca Monti. Solutions with large total variation to nonconservative hyperbolic systems. Communications on Pure & Applied Analysis, 2010, 9 (1) : 47-60. doi: 10.3934/cpaa.2010.9.47 [18] Sie Long Kek, Mohd Ismail Abd Aziz, Kok Lay Teo. A gradient algorithm for optimal control problems with model-reality differences. Numerical Algebra, Control & Optimization, 2015, 5 (3) : 251-266. doi: 10.3934/naco.2015.5.251 [19] Ruiliang Zhang, Xavier Bresson, Tony F. Chan, Xue-Cheng Tai. Four color theorem and convex relaxation for image segmentation with any number of regions. Inverse Problems & Imaging, 2013, 7 (3) : 1099-1113. doi: 10.3934/ipi.2013.7.1099 [20] Lu Tan, Ling Li, Senjian An, Zhenkuan Pan. Nonlinear diffusion based image segmentation using two fast algorithms. Mathematical Foundations of Computing, 2019, 2 (2) : 149-168. doi: 10.3934/mfc.2019011

2018 Impact Factor: 1.469