基于增强高阶非凸全变分模型的图像去噪算法

doi:10.3969/j.issn.1001-506X.2020.03.008

系统工程与电子技术 ›› 2020, Vol. 42 ›› Issue (3): 557-567.doi: 10.3969/j.issn.1001-506X.2020.03.008

基于增强高阶非凸全变分模型的图像去噪算法

刘佩¹(), 贾建^1,^2,*(), 陈莉¹(), 安影¹()

1. 西北大学信息科学与技术学院, 陕西西安 710127
2. 西北大学数学学院, 陕西西安 710127

收稿日期:2019-07-04 出版日期:2020-03-01 发布日期:2020-02-28
通讯作者: 贾建 E-mail:201720987@stumail.nwu.edu.cn;jiajian@nwu.edu.cn;chenli@nwu.edu.cn;201721013@stumail.nwu.edu.cn
作者简介:刘佩(1993-),女,硕士研究生,主要研究方向为机器学习、图像处理。E-mail:201720987@stumail.nwu.edu.cn|陈莉(1963-),女,教授,博士研究生导师,博士,主要研究方向为数据库、数据挖掘、智能信息处理。E-mail:chenli@nwu.edu.cn|安影(1995-),女,硕士研究生,主要研究方向为信息融合。E-mail:201721013@stumail.nwu.edu.cn
基金资助:
西北大学紫藤国际合作计划项目(389040008)

Image denoising algorithm based on boosting high order non-convex total variation model

Pei LIU¹(), Jian JIA^1,^2,*(), Li CHEN¹(), Ying AN¹()

1. School of Information and Technology, Northwest University, Xi'an 710127, China
2. School of Mathematics, Northwest University, Xi'an 710127, China

Received:2019-07-04 Online:2020-03-01 Published:2020-02-28
Contact: Jian JIA E-mail:201720987@stumail.nwu.edu.cn;jiajian@nwu.edu.cn;chenli@nwu.edu.cn;201721013@stumail.nwu.edu.cn
Supported by:
西北大学紫藤国际合作计划项目(389040008)

摘要/Abstract

摘要：

为了在缓解阶梯效应的同时更好地保留去噪后图像的细节信息,提出一种基于增强高阶非凸全变分(higher order non-convex total variation, HONTV)模型的图像去噪算法。该算法将每一次去噪后的图像和原始图像取平均作为增强HONTV模型下一次循环的输入并更新参数,然后采用增广拉格朗日乘子法和交替方向乘子法进行循环求解,经过多次迭代,最终得到的去噪图像包含较多的细节信息。在基于全变分的图像去噪方法中,对添加不同标准差大小的高斯白噪声的测试图像和视频进行实验。实验结果表明,所提算法在视觉性能和客观评价指标方面均优于对比算法。

关键词: 高阶非凸全变分, 图像去噪, 增广拉格朗日乘子法, 交替方向乘子法

Abstract:

In order to ease the stair casing artifacts and better preserve the details of the image after denoising, this paper proposes an image denoising algorithm based on the boosting high order non-convex total variation (HONTV) model. By averaging each denoised image and the original image as the input of the next cycle of the boosting HONTV model and updating the parameters, the augmented Lagrangian method and the alternating direction method of multipliers are used to solve the model. After multiple iterations, the resulting denoised image contains more detail information. In the image denoising method based on total variation, the experimental results show that the proposed algorithm outperforms the comparison algorithm in terms of visual performance and objective evaluation index for adding Gaussian white noise test images and videos with different standard deviations.

Key words: high order non-convex total variation (HONTV), image denoising, augmented Lagrangian method, alternating direction method of multipliers

中图分类号:

TP391.4

刘佩, 贾建, 陈莉, 安影. 基于增强高阶非凸全变分模型的图像去噪算法[J]. 系统工程与电子技术, 2020, 42(3): 557-567.

Pei LIU, Jian JIA, Li CHEN, Ying AN. Image denoising algorithm based on boosting high order non-convex total variation model[J]. Systems Engineering and Electronics, 2020, 42(3): 557-567.

图/表 16

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

图13

图14

表1

表2

参考文献 34

1	RUDIN L I , OSHER S , FATEMI E . Nonlinear total variation based noise removal algorithms[J]. Physica D:nonlinear phenomena, 1992, 60 (1/4): 259- 268.
2	GOLDSTEIN T , OSHER S . The split Bregman method for L₁-regularized problems[J]. Siam urnal on imaging sciences, 2009, 2 (2): 323- 343. doi: 10.1137/080725891
3	MICCHELLI C A, SHEN L, XU Y. Proximity algorithms for image models: denoising[J]. Inverse Problems, 2011, 27(4): Article ID: 045009.
4	YOU Y L , KAVEH M . Fourth-order partial differential equations for noise removal[J]. IEEE Trans.on Image Processing, 2000, 9 (10): 1723- 1730. doi: 10.1109/83.869184
5	LYSAKER M , LUNDERVOLD A , TAI X C . Noise removal using fourth-order partial differential equation with applications to medical magnetic resonance images in space and time[J]. IEEE Trans.on Image Processing, 2003, 12 (12): 1579- 1590. doi: 10.1109/TIP.2003.819229
6	LV X G , SONG Y Z , WANG S X , et al. Image restoration with a high-order total variation minimization method[J]. Applied Mathematical Modelling, 2013, 37 (16/17): 8210- 8224.
7	BREDIES K , KUNISCH K , POCK T . Total generalized variation[J]. SIAM Journal on Imaging Sciences, 2010, 3 (3): 492- 526. doi: 10.1137/090769521
8	HE C H , HU C H , YANG X G , et al. An adaptive total generalized variation model with augmented Lagrangian method for image denoising[J]. Mathematical Problems in Engineering, 2014. doi: 10.1155/2014/157893
9	WALI S , ZHANG H , CHANG H , et al. A new adaptive boosting total generalized variation (TGV) technique for image denoising and inpainting[J]. Journal of Visual Communication and Image Representation, 2019, 59, 39- 51. doi: 10.1016/j.jvcir.2018.12.047
10	PAPAFITSOROS K , SCHÖNLIEB C B . A combined first and second order variational approach for image reconstruction[J]. Journal of Mathematical Imaging and Vision, 2014, 48 (2): 308- 338. doi: 10.1007/s10851-013-0445-4
11	LIU J , HUANG T Z , SELESNICK I W , et al. Image restoration using total variation with overlapping group sparsity[J]. Information Sciences, 2015, 295 (C): 232- 246.
12	LIU R W, WU D, WU C S, et al. Constrained nonconvex hybrid variational model for edge-preserving image restoration[C]//Proc.of the IEEE International Conference on Systems, Man, and Cybernetics, 2015: 1809-1814.
13	SHI M , HAN T , LIU S . Total variation image restoration using hyper-Laplacian prior with overlapping group sparsity[J]. Signal Processing, 2016, 126, 65- 76. doi: 10.1016/j.sigpro.2015.11.022
14	WANG H Y , LI Y , CEN Y G , et al. Multi-matrices low-rank decomposition with structural smoothness for image denoising[J]. IEEE Trans.on Circuits and Systems for Video Technology, 2019. doi: 10.1109/TCSVT.2019.2890880
15	WANG H , CEN Y , HE Z , et al. Reweighted low-rank matrix analysis with structural smoothness for image denoising[J]. IEEE Trans.on Image Processing, 2018, 27 (4): 1777- 1792. doi: 10.1109/TIP.2017.2781425
16	KUMAR A , AHMAD M O , SWAMY M N S . Tchebichef and adaptive steerable-based total variation model for image denoising[J]. IEEE Trans.on Image Processing, 2019, 28 (6): 2921- 2935. doi: 10.1109/TIP.2019.2892663
17	ADAM T , PARAMESRAN R . Image denoising using combined higher order non-convex total variation with overlapping group sparsity[J]. Multidimensional Systems and Signal Processing, 2019, 30 (1): 503- 527. doi: 10.1007/s11045-018-0567-3
18	SELESNICK I W, CHEN P Y. Total variation denoising with overlapping group sparsity[C]//Proc.of the IEEE International Conference on Acoustics, Speech and Signal Processing, 2013: 5696-5700.
19	PEYRÉ G, FADILI J. Group sparsity with overlapping partition functions[C]//Proc.of the 19th IEEE European Signal Processing Conference, 2011: 303-307.
20	BOYD S , PARIKH N , CHU E , et al. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends in Machine learning, 2011, 3 (1): 1- 122.
21	CANDES E J , WAKIN M B , BOYD S P . Enhancing sparsity by reweighted $\ell$₁ minimization[J]. Journal of Fourier Analysis and Applications, 2008, 14 (5/6): 877- 905.
22	LYU Q , LIN Z , SHE Y , et al. A comparison of typical $\ell$_p minimization algorithms[J]. Neurocomputing, 2013, 119 (16): 413- 424.
23	CHEN X , ZHOU W . Convergence of the reweighted $\ell$₁ minimization algorithm for $\ell$₂-$\ell$_p minimization[J]. Computational Optimization and Applications, 2014, 59 (1/2): 47- 61.
24	CHAN R H , TAO M , YUAN X . Constrained total variation deblurring models and fast algorithms based on alternating direction method of multipliers[J]. SIAM Journal on imaging Sciences, 2013, 6 (1): 680- 697. doi: 10.1137/110860185
25	HE C , HU C H , LI X D , et al. A parallel alternating direction method with application to compound $\ell$₁-regularized imaging inverse problems[J]. Information Sciences, 2016, 348, 179- 197. doi: 10.1016/j.ins.2016.01.087
26	WANG Y , YIN W , ZENG J . Global convergence of ADMM in nonconvex nonsmooth optimization[J]. Journal of Scientific Computing, 2019, 78 (1): 29- 63.
27	TAO M, YANG J F. Alternating direction algorithms for total variation deconvolution in image reconstruction[R]. Nanjing University, 2009.
28	CONDAT L . A generic proximal algorithm for convex optimization—application to total variation minimization[J]. IEEE Signal Processing Letters, 2014, 21 (8): 985- 989. doi: 10.1109/LSP.2014.2322123
29	UNIVERSITY of Granada. Standard denoised image data set.[EB/OL].[2019-05-06].http://decsai.ugr.es/~javier/denoise/test-images.
30	JIAN Z. The source code of the SBATV method.[EB/OL].[2019-03-01].https://github.com/zj15001/nonconvex_TV_denoise.
31	TARMIZI ADAM. The source code of the PATV method.[EB/OL].[2011-03-17]. https://github.com/tarmiziAdam2005.
32	TARMIZI ADAM. The source code of the OGSTV method.[EB/OL].[2015-09-22]. https://github.com/tarmiziAdam2005.
33	TARMIZI ADAM. The source code of the HONTV method.[EB/OL].[2018-03-17]. https://github.com/tarmiziAdam2005.
34	WANG Z , BOVIK A C , SHEIKH H R , et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans.on Image Processing, 2004, 13 (4): 600- 612. doi: 10.1109/TIP.2003.819861

帧数	SBATV	PATV	OGSTV	HONTV	本文算法
86	28.63/0.853	28.90/0.853	28.89/0.834	29.32/0.876	29.50/0.877
50	28.63/0.857	29.35/0.857	29.09/0.824	29.74/0.878	29.86/0.882
49	28.68/0.853	29.40/0.859	29.10/0.827	29.59/0.876	29.86/0.880
53	28.59/0.855	29.26/0.858	29.02/0.835	29.44/0.874	29.80/0.881
4	29.09/0.865	29.75/0.866	29.35/0.830	29.80/0.885	30.19/0.888
18	28.68/0.859	29.05/0.860	28.87/0.832	29.48/0.878	29.65/0.880
28	28.47/0.858	29.14/0.861	28.84/0.835	29.42/0.879	29.87/0.883
54	28.74/0.858	29.10/0.856	28.99/0.824	29.38/0.875	29.65/0.878
58	28.39/0.853	29.00/0.862	28.71/0.831	29.43/0.875	29.54/0.878
23	28.56/0.856	29.05/0.859	28.83/0.826	29.44/0.879	29.68/0.882
25	28.61/0.858	29.06/0.862	28.85/0.830	29.51/0.881	29.55/0.882
24	28.53/0.859	29.04/0.859	28.89/0.828	29.62/0.883	29.67/0.883
88	28.57/0.857	28.99/0.855	28.89/0.839	29.37/0.882	29.54/0.880
5	29.22/0.866	29.58/0.859	29.27/0.831	29.72/0.881	30.05/0.882
31	28.68/0.860	29.27/0.865	28.73/0.826	29.24/0.876	29.77/0.885
36	28.67/0.850	29.28/0.858	28.83/0.820	29.49/0.876	29.76/0.877
15	28.76/0.865	29.26/0.863	28.83/0.829	29.50/0.880	29.86/0.884
37	28.62/0.852	29.36/0.859	29.01/0.824	29.39/0.873	29.61/0.876
74	29.12/0.857	29.71/0.861	29.49/0.825	29.72/0.877	30.05/0.882
89	28.74/0.860	29.07/0.859	28.96/0.835	29.51/0.875	29.65/0.879
平均值	28.69/0.857	29.23/0.859	28.97/0.829	29.50/0.878	29.75/0.880

帧数	SBATV	PATV	OGSTV	HONTV	本文算法
70	25.10/0.731	25.23/0.732	25.14/0.726	25.23/0.754	25.33/0.763
87	24.24/0.721	24.54/0.717	24.61/0.725	24.67/0.742	24.60/0.753
47	24.90/0.744	24.80/0.737	24.73/0.723	24.98/0.749	25.05/0.760
79	25.05/0.733	24.99/0.713	24.76/0.715	24.96/0.741	25.19/0.755
90	24.67/0.740	24.65/0.721	24.76/0.731	24.94/0.748	24.84/0.757
85	24.63/0.726	24.57/0.719	24.49/0.714	24.57/0.738	24.52/0.736
94	25.12/0.737	24.89/0.729	24.92/0.723	24.73/0.750	25.04/0.754
30	24.60/0.744	24.73/0.744	24.69/0.715	24.65/0.753	24.95/0.764
39	24.76/0.751	24.85/0.740	24.77/0.733	24.81/0.740	24.69/0.755
26	24.61/0.746	24.85/0.740	24.44/0.728	24.65/0.751	24.94/0.770
41	24.95/0.752	24.86/0.740	24.96/0.735	24.82/0.744	25.15/0.763
81	24.64/0.720	24.88/0.821	24.88/0.716	24.86/0.745	24.97/0.750
42	25.10/0.749	24.89/0.736	24.91/0.726	24.91/0.747	25.15/0.763
54	24.54/0.725	24.50/0.726	24.53/0.723	24.62/0.741	24.69/0.743
10	24.89/0.738	24.78/0.734	24.84/0.731	24.93/0.756	25.14/0.766
51	24.74/0.735	25.03/0.730	24.77/0.725	24.89/0.749	25.13/0.759
34	24.81/0.735	24.72/0.742	24.53/0.724	24.76/0.746	24.84/0.751
18	24.63/0.735	24.73/0.743	24.63/0.726	24.74/0.739	24.66/0.748
55	24.40/0.727	24.50/0.728	24.43/0.718	24.65/0.737	24.86/0.752
4	24.90/0.747	24.88/0.738	24.87/0.726	24.87/0.743	25.06/0.757
平均值	24.78/0.738	24.85/0.731	24.70/0.724	24.81/0.746	24.98/0.755

[1]	陈玉香, 陈胜垚, 冉龙瑶, 席峰. 动态超表面天线阵列混合发射波束成形[J]. 系统工程与电子技术, 2022, 44(3): 730-736.
[2]	杨磊, 张苏, 盖明慧, 方澄. 高分辨SAR目标成像方向性结构特征增强[J]. 系统工程与电子技术, 2022, 44(3): 808-818.
[3]	陈诚, 刘涛, 宋晓骥, 粟毅, 金光虎. 基于稀疏信号处理的穿透成像增强算法[J]. 系统工程与电子技术, 2021, 43(8): 2021-2027.
[4]	丁逊, 张劲东, 王娜, 王玉莹. 基于相参积累的捷变频雷达系统相位误差估计与稀疏场景重构算法[J]. 系统工程与电子技术, 2021, 43(6): 1515-1523.
[5]	邵凯, 陈连成, 刘胤. 高移动性Jakes信道的学习与估计[J]. 系统工程与电子技术, 2021, 43(4): 1119-1125.
[6]	杨真真, 乐俊, 杨永鹏, 范露. 非凸运动辅助低秩稀疏分解目标检测算法[J]. 系统工程与电子技术, 2020, 42(6): 1218-1225.
[7]	王彩云, 赵焕玥, 王佳宁, 李晓飞, 黄盼盼. 快速加权核范数最小化的SAR图像去噪算法[J]. 系统工程与电子技术, 2019, 41(7): 1504-1508.
[8]	王彩云, 胡允侃, 吴淑侠. 基于贝叶斯模型的shearlet域SAR图像去噪方法[J]. 系统工程与电子技术, 2017, 39(6): 1250-1255.
[9]	胡悦, 仲崇潇, 曹梦宇, 赵旷世. 基于增广拉格朗日乘子的快速高阶全变分图像去噪方法[J]. 系统工程与电子技术, 2017, 39(12): 2831-2839.
[10]	李博, 谢巍. 基于自适应分数阶微积分的图像去噪与增强算法[J]. 系统工程与电子技术, 2016, 38(1): 185-192.
[11]	刘书君, 吴国庆, 张新征, 徐礼培. 基于Shearlet域系数处理的SAR图像降噪[J]. 系统工程与电子技术, 2015, 37(9): 2023-2028.
[12]	黄果, 许黎, 陈庆利, 蒲亦非. 基于时间-空间分数阶偏微分方程的图像去噪模型[J]. Journal of Systems Engineering and Electronics, 2012, 34(8): 1741-1752.
[13]	刘帅奇，胡绍海，肖扬. 基于局部混合滤波的SAR图像去噪[J]. Journal of Systems Engineering and Electronics, 2012, 34(2): 396-402.
[14]	胡根生1,2. 梁栋1,2，黄林生1,2. 基于支持向量值轮廓波变换的遥感图像去噪[J]. Journal of Systems Engineering and Electronics, 2011, 33(7): 1658-1663.
[15]	黄果，蒲亦非，陈庆利，周激流 . 基于分数阶积分的图像去噪[J]. Journal of Systems Engineering and Electronics, 2011, 33(4): 925-932.

基于增强高阶非凸全变分模型的图像去噪算法

Image denoising algorithm based on boosting high order non-convex total variation model

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 16

参考文献 34

相关文章 15

编辑推荐

Metrics

本文评价