基于改进的自适应差分演化算法的二维Otsu多阈值图像分割

罗钧; 杨永松; 侍宝玉

doi:10.11999/JEIT180949

基于改进的自适应差分演化算法的二维Otsu多阈值图像分割

doi: 10.11999/JEIT180949

重庆大学光电技术及系统教育部重点实验室重庆 400030

详细信息

作者简介:
罗钧：男，1963年生，教授，博士生导师，研究方向为模式识别与人工智能，精密机械及测试计量，智能信息处理

杨永松：男，1994年生，硕士生，研究方向为嵌入式系统，机器视觉

侍宝玉：女，1994年生，硕士生，研究方向为嵌入式系统，机器视觉

通讯作者:
罗钧　luojun@cqu.edu.cn

中图分类号: TP391.41
计量
- 文章访问数: 3629
- HTML全文浏览量: 1360
- PDF下载量: 177
- 被引次数: 0
出版历程
- 收稿日期: 2018-10-12
- 修回日期: 2019-03-04
- 网络出版日期: 2019-03-28
- 刊出日期: 2019-08-01

Multi-threshold Image Segmentation of 2D Otsu Based on Improved Adaptive Differential Evolution Algorithm

Key Laboratory of Optoelectronic Technology and System of Ministry of Education, Chongqing University, Chongqing 400030, China

摘要

摘要: 针对常规最大类间方差法在多阈值图像分割中存在的运算量大、计算时间长、分割精度较低等问题，该文提出一种基于改进的自适应差分演化(JADE)算法的2维Otsu多阈值分割法。首先，为增强初始化种群的质量、提升控制参数的适应性，将混沌映射机制融入到JADE算法中；进而，通过该改进算法求解2维 Otsu 多阈值图像的最佳分割阈值；最终，将该算法与差分进化(DE), JADE，改进正弦参数自适应的差分进化(LSHADE-cnEpSin)以及增强的适应性微分变换差分进化(EFADE) 4种算法的2维Otsu多阈值图像分割进行比较。实验结果表明，与其它4种算法相比，基于改进JADE算法的2维Otsu多阈值图像分割在分割速度以及精度上均有较明显的改善。
- 图像分割 /
- 最大类间方差法 /
- 混沌映射 /
- 改进的自适应差分演化算法
Abstract: The multi-threshold image segmentation of the classical 2D maximal between-cluster variance method has deficiencies such as large computation, long calculation time, low segmentation precision and so on. A multi-threshold segmentation of 2D Otsu based on improved Adaptive Differential Evolution (JADE) algorithm is proposed. Firstly, in order to enhance the quality of the initialized population and improve the adaptability of the control parameters, the chaotic mapping mechanism is integrated into the JADE algorithm. Furthermore, the optimal segmentation threshold of 2D Otsu multi-threshold image is solved by improved JADE algorithm. Finally, the algorithm is compared with multi-threshold image segmentation method of 2D Otsu based on Differential Evolution (DE), JADE, Improved Differential Evolution with Adaptive Sinusoidal Parameters (LSHADE-cnEpSin) and Enhanced Adaptive Differential Transformation Differential Evolution (EFADE) algorithm. The experimental results show that compared with the other four algorithms, the multi-threshold image segmentation of 2D Otsu based on the improved JADE algorithm has a significant improvement in terms of segmentation speed and accuracy.
- Image segmentation /
- Maximum interclass variance method /
- Chaotic map /
- Improved Adaptive Differential Evolution (JADE) algorithm

HTML全文

图 1 2维多阈值分割直方图

下载: 全尺寸图片幻灯片

图 2 基于CJADE算法2维Otsu多阈值分割方法流程图

下载: 全尺寸图片幻灯片

图 3 分割效果图

下载: 全尺寸图片幻灯片

图 4 进化曲线图

下载: 全尺寸图片幻灯片

表 1 算法1：混沌映射更新参数u_F和u_CR的伪代码

　(1) If $\;\alpha < \beta $

　(2) 　　${u_{\rm CR}} = {u_1} \cdot {u_{\rm CR}} \cdot (1 - {u_{\rm CR}})$

　(3) 　　${u_F} = {u_2} \cdot {u_F} \cdot (1 - {u_F})$

　(4) Else

　(5) 　　${u_{\rm CR}} = (1 - c) \cdot {u_{\rm CR}} + c \cdot {{\rm mean}_{\rm A}}({S_{\rm CR}})$

　(6) 　　${u_F} = (1 - c) \cdot {u_F} + c \cdot {{\rm mean}_{\rm L}}({S_F})$

　(7) End If

下载: 导出CSV

表 2 PSNR、运算时间以及迭代次数的对比

算法		Lena (512$ \times $512)			Finger (256$ \times $256)			Pepper (512$ \times $512)
算法		2阈值	3阈值	4阈值	2阈值	3阈值	4阈值	2阈值	3阈值	4阈值
DE算法	PSNR(dB)	10.58	13.88	15.64	12.02	12.45	14.14	11.68	15.84	16.54
	收敛时间(s)	7.79	7.82	7.84	3.64	3.58	3.73	8.49	8.34	8.82
	迭代次数	72	58	64	62	57	66	45	43	47
JADE算法	PSNR(dB)	11.79	14.25	16.02	12.35	13.02	14.26	11.71	16.32	16.71
	收敛时间(s)	0.85	0.83	0.77	0.51	0.53	0.57	0.81	0.80	0.83
	迭代次数	52	54	50	59	62	58	60	56	58
LSHADE-cnEpSin算法	PSNR(dB)	13.70	14.98	15.67	12.07	12.77	14.46	12.23	16.19	17.02
	收敛时间(s)	0.79	0.75	0.82	0.45	0.48	0.46	0.78	0.82	0.78
	迭代次数	34	35	33	65	45	60	50	48	46
EFADE算法	PSNR(dB)	12.89	15.05	15.45	13.23	12.61	13.24	12.11	15.57	16.67
	收敛时间(s)	0.99	1.12	1.10	0.77	0.76	0.83	1.24	1.31	1.29
	迭代次数	45	42	46	50	48	52	40	38	41
CJADE算法	PSNR(dB)	13.93	15.64	16.25	13.65	14.67	14.89	12.56	16.57	17.12
	收敛时间(s)	0.64	0.66	0.65	0.45	0.44	0.48	0.61	0.64	0.66
	迭代次数	38	35	38	41	40	44	40	36	38

下载: 导出CSV

表 3 阈值和距离测度值的对比

算法		Lena (512$ \times $512)			Finger (256$ \times $256)			Pepper (512$ \times $512)
算法		2阈值	3阈值	4阈值	2阈值	3阈值	4阈值	2阈值	3阈值	4阈值
DE算法	距离测度	4645.67	4698.86	4747.74	1223.45	1247.75	1296.25	5340.87	5407.71	5513.28
DE算法	阈值	(68,71) (117,153)	(30, 32) (86,138) (193,199)	(88,95) (119,123) (151,153) (202,207)	(39,53) (155,165)	(108,124) (147,152) (168,180)	(23,38) (102,133) (150,157) (169,170)	(70,70) (117,161)	(84, 85) (142,162) (201,203)	(70,77) (111,112) (126,129) (129,179)
JADE算法	距离测度	4842.77	4912.21	4924.13	1315.43	1320.35	1326.23	5798.46	5822.86	5892.86
JADE算法	阈值	(89,149) (193,195)	(77,79) (114,149) (196,196)	(70,77) (109,137) (149,154) (182,183)	(138,166) (175,175)	(10,67) (143,164) (174,175)	(40,52) (50,110) (156,156) (171,172)	(88,91) (127,169)	(98, 115) (140,140) (178,178)	(96,101) (114,133) (149,150) (152,171)
LSHADE-cnEpSin算法	距离测度	4862.49	4905.97	4995.04	1256.65	1268.79	1289.32	5797.85	5899.34	5909.58
LSHADE-cnEpSin算法	阈值	(88,149) (194,195)	(79,79) (115,145) (177,177)	(76,76) (119,141) (158,160) (197,197)	(88,102) (183,183)	(64,82) (148,164) (183,184)	(36,39) (42,98) (145,155) (164,169)	(78,79) (126,177)	(84, 85) (127,159) (194,194)	(76,77) (121,122) (126,157) (192,193)
EFADE算法	距离测度	4848.87	4951.82	4973.23	1257.29	1267.42	1324.41	5788.61	5885.72	5892.13
EFADE算法	阈值	(89,148) (186,186)	(76,80) (130,153) (205,205)	(79,86) (112,137) (139,151) (201,203)	(44,51) (142,176)	(30,42) (140,162) (172,174)	(41,46) (68,83) (141,167) (172,173)	(86,90) (120,176)	(73, 74) (121,159) (193,194)	(53,55) (121,123) (154,154) (180,182)
CJADE算法	距离测度	4863.53	4977.34	4999.63	1327.84	1329.17	1331.28	5799.13	5898.73	5912.18
CJADE算法	阈值	(87,149) (194,194)	(77,78) (115,148) (194,195)	(78,80) (117,139) (156,156) (199,199)	(143,166) (173,173)	(25, 62) (142,166) (174,175)	(40,45) (70,98) (156,158) (162,162)	(84,85) (124,173)	(77, 78) (123,164) (195,195)	(54,55) (99,100) (129,160) (199,199)

下载: 导出CSV

参考文献(17)

刘健庄, 栗文青. 灰度图象的二维Otsu自动阈值分割法[J]. 自动化学报, 1993, 19(1): 101–105. doi: 10.16383/j.aas.1993.01.015

LIU Jianzhuang and LI Wenqing. The automatic thresholding of gray-level pictures via two-dimensional otsu method[J]. Acta Automatica Sinica, 1993, 19(1): 101–105. doi: 10.16383/j.aas.1993.01.015

申铉京, 刘翔, 陈海鹏. 基于多阈值Otsu准则的阈值分割快速计算[J]. 电子与信息学报, 2017, 39(1): 144–149. doi: 10.11999/JEIT160248

SHEN Xuanjing, LIU Xiang, and CHEN Haipeng. Fast computation of threshold based on multi-threshold Otsu criterion[J]. Journal of Electronics &Information Technology, 2017, 39(1): 144–149. doi: 10.11999/JEIT160248

HU Min, LI Mei, and WANG Ronggui. Application of an improved Otsu algorithm in image segmentation[J]. Journal of Electronic Measurement and Instrument, 2010, 24(5): 443–449. doi: 10.3724/SP.J.1187.2010.00443

ZHANG Jingqiao and SANDERSON A C. JADE: Adaptive differential evolution with optional external archive[J]. IEEE Transactions on Evolutionary Computation, 2009, 13(5): 945–958. doi: 10.1109/TEVC.2009.2014613

TANABE R and FUKUNAGA A. Success-history based parameter adaptation for differential evolution[C]. 2013 IEEE Congress on Evolutionary Computation, Cancun, Mexico, 2013: 71–78.

TANABE R and FUKUNAGA A S. Improving the search performance of SHADE using linear population size reduction[C]. 2014 IEEE Congress on Evolutionary Computation, Beijing, China, 2014: 1658–1665.

AWAD N H, ALI M Z, SUGANTHAN P N, et al. An ensemble sinusoidal parameter adaptation incorporated with L-SHADE for solving CEC2014 benchmark problems[C]. 2016 IEEE Congress on Evolutionary Computation, Vancouver, Canada, 2016: 2958–2965.

AWAD N H, ALI M Z, and SUGANTHAN P N. Ensemble sinusoidal differential covariance matrix adaptation with Euclidean neighborhood for solving CEC2017 benchmark problems[C]. 2017 IEEE Congress on Evolutionary Computation, San Sebastian, Spain, 2017: 372–379.

MOHAMED A W and SUGANTHAN P N. Real-parameter unconstrained optimization based on enhanced fitness-adaptive differential evolution algorithm with novel mutation[J]. Soft Computing, 2018, 22(10): 3215–3235. doi: 10.1007/s00500-017-2777-2

STHITPATTANAPONGSA P and SRINARK T. A two-stage Otsu’S thresholding based method on a 2D histogram[C]. IEEE 7th International Conference on Intelligent Computer Communication and Processing, Cluj-Napoca, Romania, 2011: 345–348.

张春美, 陈杰, 辛斌. 参数适应性分布式差分进化算法[J]. 控制与决策, 2014, 29(4): 701–706. doi: 10.13195/j.kzyjc.2013.0080

ZHANG Chunmei, CHEN Jie, and XIN Bin. Distributed differential evolution algorithm with adaptive parameters[J]. Control and Decision, 2014, 29(4): 701–706. doi: 10.13195/j.kzyjc.2013.0080

王李进, 钟一文, 尹义龙. 带外部存档的正交交叉布谷鸟搜索算法[J]. 计算机研究与发展, 2015, 52(11): 2496–2507. doi: 10.7544/issn1000-1239.2015.20148042

WANG Lijin, ZHONG Yiwen, and YIN Yilong. Orthogonal crossover cuckoo search algorithm with external archive[J]. Journal of Computer Research and Development, 2015, 52(11): 2496–2507. doi: 10.7544/issn1000-1239.2015.20148042

RERE L M R, FANANY M I, and MURNI A. Adaptive DE based on chaotic sequences and random adjustment for image contrast enhancement[C]. 2014 International Conference of Advanced Informatics: Concept, Theory and Application, Bandung, Indonesia, 2015: 220–225. doi: 10.1109/ICAICTA.2014.7005944.

陈志刚, 梁涤青, 邓小鸿, 等. Logistic混沌映射性能分析与改进[J]. 电子与信息学报, 2016, 38(6): 1547–1551. doi: 10.11999/JEIT151039

CHEN Zhigang, LIANG Diqing, DENG Xiaohong, et al. Performance analysis and improvement of logistic chaotic mapping[J]. Journal of Electronics &Information Technology, 2016, 38(6): 1547–1551. doi: 10.11999/JEIT151039

陈如清. 采用新型粒子群算法的电力电子装置在线故障诊断方法[J]. 中国电机工程学报, 2008, 28(24): 70–74. doi: 10.3321/j.issn:0258-8013.2008.24.012

CHEN Ruqing. A novel PSO based on-line fault diagnosis approach for power electronic system[J]. Proceedings of the CSEE, 2008, 28(24): 70–74. doi: 10.3321/j.issn:0258-8013.2008.24.012

SHA Chunshi, HOU Jian, and CUI Hongxia. A robust 2D Otsu’s thresholding method in image segmentation[J]. Journal of Visual Communication and Image Representation, 2016, 41: 339–351. doi: 10.1016/j.jvcir.2016.10.013

HUYNH-THU Q and GHANBARI M. The accuracy of PSNR in predicting video quality for different video scenes and frame rates[J]. Telecommunication Systems, 2012, 49(1): 35–48. doi: 10.1007/s11235-010-9351-x

施引文献

资源附件(0)

访问统计