基于改进麻雀搜索算法的多阈值图像分割

doi:10.12305/j.issn.1001-506X.2021.02.05

摘要/Abstract

摘要：

针对传统多阈值图像分割方法中存在的分割精度低、计算量大、分割速度慢等问题,提出了一种基于改进麻雀搜索算法(improved sparrow search algorithm, ISSA)的多阈值图像分割方法。首先,结合鸟群算法(bird swarm algorithm, BSA)中飞行行为的思想优化麻雀搜索算法(sparrow search algorithm, SSA),并采用4种类型的基准函数评估ISSA的寻优性能。然后,进行基于类间方差和Kapur熵的多阈值图像分割,并对比两种方法的分割结果。最后,采用PSNR、目标函数值和标准差作为评估标准,将ISSA与现有分割算法进行对比分析。结果表明, ISSA具有更优的搜索能力和开拓能力,且分割速度和分割精度均得到提升。

关键词: 图像分割, 改进麻雀搜索算法, 多阈值, 最大类间方差, Kapur熵

Abstract:

To solve the problems of low segmentation accuracy, large calculation amount and slow segmentation speed in the traditional multi-threshold image segmentation methods, a multi-threshold image segmentation method based on improved sparrow search algorithm (ISSA) is proposed. First, the sparrow search algorithm (SSA) is optimized based on the flight behavior in the bird swarm algorithm (BSA), and four types of benchmark functions are used to evaluate the optimization performance of ISSA. Then, the between-class variance and Kapur entropy are used to perform the multi-threshold image segmentation, and the segmentation results of the two methods are compared. Finally, using PSNR, objective function value and standard deviation as evaluation criteria, ISSA is compared with the existing segmentation algorithms. The results show that ISSA has better search ability and development ability, and it has a significant improvement in terms of segmentation speed and accuracy.

Key words: image segmentation, improved sparrow search algorithm (ISSA), multi-threshold, Otsu, Kapur entropy

中图分类号:

TP391.41

吕鑫, 慕晓冬, 张钧. 基于改进麻雀搜索算法的多阈值图像分割[J]. 系统工程与电子技术, 2021, 43(2): 318-327.

Xin LYU, Xiaodong MU, Jun ZHANG. Multi-threshold image segmentation based on improved sparrow search algorithm[J]. Systems Engineering and Electronics, 2021, 43(2): 318-327.

图/表 13

图1

图2

表1

图3

表2

表3

表4

图4

图5

表5

表6

表7

图6

参考文献 15

1	刘仲民, 王阳, 李战明, 等. 基于简单线性迭代聚类和快速最近邻区域合并的图像分割算法[J]. 吉林大学学报(工学版), 2018, 48 (6): 1931- 1937.
	LIU Z M , WANG Y , LI Z M , et al. Image segmentation algorithm based on SLIC and fast nearest neighbor region merging[J]. Journal of Jilin University(Engineering and Technology Edition), 2018, 48 (6): 1931- 1937.
2	AKAY B . A study on particle swarm optimization and artificial bee colony algorithms for multilevel thresholding[J]. Applied Soft Computing, 2013, 13 (6): 3066- 3091. doi: 10.1016/j.asoc.2012.03.072
3	YAO X T , LI Z Y , LIU L , et al. Multi-threshold image segmentation based on improved grey wolf optimization algorithm[J]. IOP Conference Series Earth and Environmental Science, 2019, 252 (4): 042105.
4	WANG S K , JIA H M , PENG X X . Modified salp swarm algorithm based multilevel thresholding for color i mage segmentation[J]. Mathematical Bioences and Engineering, 2019, 17 (1): 700- 724.
5	SATHYA P D , KAYALVIZHI R . Optimal multilevel thresholding using bacterial foraging algorithm[J]. Expert Systems with Applications, 2011, 38 (12): 15549- 15564. doi: 10.1016/j.eswa.2011.06.004
6	LU Y T , ZHAO W L , MAO X B . Multi-threshold image segmentation based on improved particle swarm optimization and maximum entropy method[J]. Advanced Materials Research, 2014, 989 (994): 3649- 3653.
7	CUEVAS E , FELIPE S , ZALDIVAR D , et al. A multi-threshold segmentation approach based on Artificial Bee Colony optimization[J]. Applied Intelligence, 2012, 37 (3): 321- 336. doi: 10.1007/s10489-011-0330-z
8	QIN J , SHEN X J , MEI F , et al. An Otsu multi-thresholds segmentation algorithm based on improved ACO[J]. The Journal of Supercomputing, 2019, 75 (2): 955- 967. doi: 10.1007/s11227-018-2622-0
9	DEHSHIBI M M , SOURIZAEI M , FAZLALI M , et al. A hybrid bio-inspired learning algorithm for image segmentation using multilevel thresholding[J]. Multimedia Tools & Applications, 2016, 76 (14): 15951- 15986.
10	XUE J K , SHEN B . A novel swarm intelligence optimization approach: sparrow search algorithm[J]. Systems Science & Control Engneering, 2020, 8 (1): 22- 34.
11	MENG X B , GAO X Z , LU L , et al. A new bio-inspired optimisation algorithm: bird swarm algorithm[J]. Journal of Experimental & Theoretical Artificial Intelligence, 2016, 28 (4): 673- 687.
12	OSTU N . A threshold selection method from gray-histogram[J]. IEEE Trans.on Systems, Man, and Cybernetics, 1978, 9 (1): 62- 66.
13	WU Y , ZHOU Y , SAVERIADES G , et al. Local Shannon entropy measure with statistical tests for image randomness[J]. Information Sciences, 2013, 222, 323- 342. doi: 10.1016/j.ins.2012.07.049
14	罗钧, 刘建强, 庞亚男. 基于邻域搜索JADE的二维Otsu多阈值图像分割[J]. 系统工程与电子技术, 2020, 42 (10): 2164- 2171. doi: 10.3969/j.issn.1001-506X.2020.10.03
	LUO J , LIU J Q , PANG Y N . Multi-threshold image segmen tation of 2D Otsu based on neighborhoods search JADE[J]. Systems Engineering and Electronics, 2020, 42 (10): 2164- 2171. doi: 10.3969/j.issn.1001-506X.2020.10.03
15	OLIVA D , CUEVAS E , PAJARES G , et al. Multilevel thresholding segmentation based on harmony search optimization[J]. Journal of Applied Mathematics, 2013, 575414.

函数	特征	函数表达式	上下界	维数
F1	US	$f_{1}(x)=\sum\limits_{i=1}^{n} x_{i}^{2}$	[-100, 100]	30
F2	UN	$f_{2}(x)=\sum\limits_{i=1}^{n-1}\left[100\left(x_{i+1}-x_{i}^{2}\right)+\left(x_{i}-1\right)^{2}\right]$	[-30, 30]	30
F3	MS	$\begin{aligned} f_{3}(x)=& 0.1\left\{\sin ^{2}\left(3 \pi x_{1}\right)+\sum\limits_{i=1}^{n}\left(x_{i}-1\right)^{2}\left[1+\sin ^{2}\left(3 \pi x_{i}+1\right)\right]+\right.\\ &\left.\left(x_{n}-1\right)^{2}\left[1+\sin ^{2}\left(2 \pi x_{n}\right)\right]\right\}+\sum\limits_{i=1}^{n} u\left(x_{i}, 5, 100, 4\right) \end{aligned}$	[-50, 50]	30
F4	MN	$f_{4}(x)=-\sum\limits_{i=1}^{n} \sin \left(x_{i}\right)\left(\sin \left(i x_{i}^{2} \div \pi\right)\right)^{20}$	[0, π]	10

函数	算法	最劣值	最优值	平均值	标准差	平均运行时间
F1	SSA	5.377 9E-39	0	2.423 6E-40	1.029 6E-39	0.351 0
F1	ISSA	7.114 7E-47	1.673 9E-56	3.284 1E-48	1.323 9E-47	0.308 3
F2	SSA	0.022 9	1.051 4E-07	0.003 0	0.006 0	0.442 2
F2	ISSA	0.022 6	2.521 3E-09	0.002 4	0.004 7	0.402 6
F3	SSA	1.657 7E-04	3.172 8E-08	1.506 2E-05	3.243 6E-05	0.849 5
F3	ISSA	4.757 6E-05	3.120 1E-11	7.826 6E-06	1.160 2E-05	0.827 1
F4	SSA	-5.516 0	-9.339 3	-7.456 0	0.850 0	0.378 1
F4	ISSA	-5.923 3	-9.483 9	-8.109 9	0.819 5	0.348 4

图像	n	阈值	PSNR	STD
Cameraman	2	70 144	17.249 1	1.348 7E-06
	3	59 119 156	20.216 5	0.033 9
	4	4 295 140 170	21.526 1	0.120 6
	5	3 682 122 149 173	23.286 2	0.112 9
Lena	2	92 151	15.458 5	0
	3	80 126 170	17.490 5	0
	4	74 113 145 180	18.883 7	1.168 0E-06
	5	73 109 136 160 188	20.267 2	0.791 6
Butterfly	2	85 148	14.726 2	0
	3	74 120 171	16.994 1	1.168 0E-06
	4	6 698 134 176	18.705 8	0.016 6
	5	6 288 113 143 180	19.747 3	0.097 1
Baboon	2	97 149	15.425 3	0
	3	85 125 161	17.718 0	0.032 3
	4	72 106 137 168	20.707 5	0.042 0
	5	68 100 126 150 175	21.438 9	0.063 8

图像	n	阈值	PSNR	STD
Cameraman	2	128 193	13.647 5	0.000 4
	3	44 104 193	14.609 5	0.032 2
	4	4 497 146 197	20.209 0	0.001 7
	5	2 562 100 146 197	20.806 9	0.062 9
Lena	2	97 164	14.580 6	0
	3	2 497 164	16.226 1	0.139 4
	4	2 482 126 175	19.276 6	0.212 4
	5	246 497 137 179	20.938 7	0.278 3
Butterfly	2	97 157	14.522 4	7.105 4E-15
	3	84 139 205	15.017 6	0.000 1
	4	72 114 156 205	17.408 2	0.000 4
	5	1 672 114 156 205	18.861 0	0.040 7
Baboon	2	79 143	16.018 3	1.065 8E-14
	3	4 699 152	18.604 7	0.000 7
	4	3 374 114 159	20.501 8	0.002 3
	5	3 369 104 138 172	22.371 0	0.004 9

图像	n	P-value Otsu versus Kapur
Cameraman	2	6.115 1E-14
	3	8.467 6E-13
	4	3.038 0E-12
	5	1.959 6E-11
Lena	2	1.685 3E-14
	3	5.297 9E-13
	4	2.270 6E-11
	5	1.251 3E-07
Butterfly	2	1.685 3E-14
	3	1.333 4E-13
	4	1.570 2E-13
	5	7.290 6E-12
Baboon	2	1.685 3E-14
	3	7.544 2E-13
	4	1.385 6E-12
	5	2.036 8E-12