基于GEMD与改进PCNN的红外与可见光图像融合

杨艳春; 李小苗; 党建武; 王阳萍

doi:10.13700/j.bh.1001-5965.2022.0756

基于GEMD与改进PCNN的红外与可见光图像融合

doi: 10.13700/j.bh.1001-5965.2022.0756

兰州交通大学电子与信息工程学院，兰州 730070

基金项目: 长江学者和创新团队发展计划资助(IRT_16R36)；国家自然科学基金(62067006)；甘肃省科技计划项目(18JR3RA104)；甘肃省高等学校产业支撑计划项目(2020C-19)；兰州市科技计划项目(2019-4-49)；甘肃省教育厅青年博士基金(2022QB-067)；甘肃省自然科学基金(23JRRA847, 21JR7RA300)；兰州交通大学天佑创新团队(TY202003)；兰州交通大学—天津大学联合创新基金(2021052)

详细信息

通讯作者:
E-mail：yangyanchun102@ sina.com

中图分类号: TP391
计量
- 文章访问数: 2044
- HTML全文浏览量: 118
- PDF下载量: 32
- 被引次数: 0
出版历程
- 收稿日期: 2022-09-02
- 录用日期: 2023-01-02
- 网络出版日期: 2023-02-06
- 整期出版日期: 2023-10-01

Infrared and visible image fusion based on GEMD and improved PCNN

School of Electronic and Information Engineering, Lan zhou Jiaotong University, Lan zhou 730070, China

Funds: Program for Changjiang Scholars and Innovative ResearchTeam (IRT_16R36); National Natural Science Foundation of China (62067006); Gansu Provincial Science and Technology Plan Project (18JR3RA104); Gansu Province Higher Education Industry Support Program Project (2020C-19); Lanzhou Science and Technology Plan Project (2019-4-49); Gansu Provincial Department of Education: Youth Doctoral Fund Project (2022QB-067); Natural Science Foundation of Gansu Province (23JRRA847, 21JR7RA300); Tianyou Innovation Team of Lanzhou Jiaotong University (TY202003); Lanzhou Jiaotong University-Tianjin University Joint Innovation Fund Project (2021052)

More Information

Corresponding author: E-mail：yangyanchun102@ sina.com

摘要

摘要:
针对传统图像融合方法因分解工具的局限性使融合图像边缘出现伪影与亮度、对比度下降的问题，提出一种基于梯度保边多层级分解(GEMD)与改进脉冲耦合神经网络(PCNN)的红外与可见光图像融合方法。利用梯度双边滤波器(GBF)与梯度滤波器(GF)构造一种多层级分解模型，将源图像分解为3层特征图与1个基础层，且每层特征图有细、粗2个结构；根据各特征图所包含信息特点，分别采用在输入刺激中引入改进拉普拉斯算子来增强对图像中弱细节信息捕捉的PCNN、区域能量和对比度显著的融合规则进行子图融合，得到各子特征融合图像与基础层融合图像；将各子融合图像进行叠加，获得最终融合图像。实验结果表明：所提方法在视觉效果方面与定量评价方面均有所提高，提高了红外与可见光融合图像的亮度与对比度信息。
- 图像融合 /
- 红外与可见光图像 /
- 脉冲耦合神经网络 /
- 双边滤波器 /
- 梯度滤波器
Abstract:
Because of the limitations of decomposition tools in traditional image fusion methods, artifacts, decreased brightness, and contrast appear on fused image edges. A gradient edge-preserving multi-level decomposition(GEMD)-based approach for infrared and visible image fusion, as well as an enhanced pulse-coupled neural network(PCNN), are described. The gradient bilateral filter(GBF) is proposed based on the bilateral filter and the gradient filter(GF), which can preserve the edge structure, brightness, and contrast information while smoothing the detail information. Firstly, the source images are divided into three layers of feature maps and a base layer, and then a multi-level decomposition model is built using a gradient bilateral filter and gradient filter. Each layer of feature maps has two different structures, thin and thick. Then, according to the characteristics of the information contained in each feature map, the PCNN, which introduces an improved Laplacian operator in the input stimulus to enhance weak details captured in the image, regional energy, and contrast saliency are respectively adopted to obtain the fusion images of each sub-feature and the fusion image of the base layer as the fusion rules. Finally, the sub-fusion images are superimposed to obtain the final fusion image. Through experimental verification, the proposed method has improved both visual effect and quantitative evaluation and improved the brightness and contrast information of infrared and visible fusion images.
- image fusion /
- infrared and visible image /
- pulse coupled neural network /
- bilateral filter /
- gradient filter

HTML全文

图 1 脉冲耦合神经网络模型

Figure 1. PCNN model

下载: 全尺寸图片幻灯片

图 2 梯度双边滤波器性能分析图

Figure 2. Performance analysis diagram of GBF

下载: 全尺寸图片幻灯片

图 3 梯度保边多层级分解模型

Figure 3. Gradient edge-preserving multi-level decomposition model

下载: 全尺寸图片幻灯片

图 4 差图对比图

Figure 4. Difference map comparison map

下载: 全尺寸图片幻灯片

图 5 改进PCNN性能分析

Figure 5. Improved PCNN performance analysis

下载: 全尺寸图片幻灯片

图 6 本文方法结构图

Figure 6. Structure diagram of proposed method

下载: 全尺寸图片幻灯片

图 7 分解级数分析图

Figure 7. Decomposition series analysis diagram

下载: 全尺寸图片幻灯片

图 8 权重取值分析图

Figure 8. Weight value analysis diagram

下载: 全尺寸图片幻灯片

图 9 实验源图像

Figure 9. Experimental source images

下载: 全尺寸图片幻灯片

图 10 实验结果

Figure 10. Experimental results

下载: 全尺寸图片幻灯片

图 11 各指标折线图

Figure 11. Line chart of each indicator

下载: 全尺寸图片幻灯片

表 1 实验图像各评价指标值

Table 1. The evaluation index values of experimental images

图像编号	方法	AG	IE	ID	SD	VIFF	EIR/10⁹
img1	GF	2.9693	6.7261	4.4568	40.3069	0.5405	6.7065
	RGF	4.2436	6.7333	6.2152	35.7304	0.5547	7.7050
	CNN	4.0077	7.0141	6.2199	52.0127	0.6810	10.0100
	CSMCA	3.4873	6.4559	5.1348	31.7815	0.5246	7.7815
	BRG	3.8183	5.8444	6.0292	43.1289	0.5897	1.3966
	PIAF	2.8606	6.5322	2.4365	31.9968	0.3546	7.2065
	本文方法	5.3117	7.2662	8.9986	59.4290	0.7711	11.3470
img2	GF	11.7449	7.3420	13.6928	59.6281	0.7038	7.8094
	RGF	14.3084	7.4782	17.0913	62.9388	0.7687	10.0070
	CNN	14.0207	7.3708	16.3577	68.0267	0.8093	10.1630
	CSMCA	14.1432	7.5362	16.3964	57.2800	0.7777	10.0060
	BRG	15.6231	7.4355	18.1672	65.5647	0.8082	12.3750
	PIAF	12.6485	7.2546	16.6218	60.2984	0.7548	11.6570
	本文方法	17.7200	7.4675	21.0224	75.6961	0.9057	16.5710
img3	GF	5.0065	6.2106	5.6306	31.6452	0.4731	2.1302
	RGF	6.5371	6.4985	7.5590	34.5977	0.5516	2.9668
	CNN	6.3240	6.6925	7.2309	37.3868	0.6143	2.8205
	CSMCA	5.9061	6.5271	6.6218	32.8467	0.5399	2.5836
	BRG	6.3112	6.7150	6.9639	38.2892	0.6241	2.7484
	PIAF	7.1007	6.7781	8.0320	38.3769	0.6471	3.4837
	本文方法	7.5605	6.6161	8.4529	46.2119	0.7112	4.3446
img4	GF	6.5674	7.1388	7.7132	41.8089	0.4647	3.0893
	RGF	8.5620	7.1932	10.4039	43.8182	0.5198	4.6488
	CNN	8.7445	7.4203	10.2977	51.7043	0.6275	4.8712
	CSMCA	8.9949	7.4118	10.4165	46.1752	0.4512	5.3171
	BRG	8.7851	7.4321	10.3676	48.1222	0.5135	5.0349
	PIAF	8.4339	7.3056	9.7002	49.0388	0.5860	5.0457
	本文方法	11.4676	7.6322	13.9933	63.1705	0.6450	7.9446
img5	GF	1.8168	5.3317	2.2802	15.1546	0.2358	1.0312
	RGF	2.8488	5.4740	3.5476	17.4340	0.3261	1.0513
	CNN	2.5592	6.3555	3.2363	26.9346	0.2652	1.1193
	CSMCA	2.2246	5.2983	2.7360	14.5069	0.1507	1.2474
	BRG	2.8004	5.8712	3.3013	19.9049	0.1157	1.3562
	PIAF	1.9253	5.7237	2.3440	16.5379	0.1497	1.0371
	本文方法	3.3037	5.7163	4.1779	35.7829	0.3438	1.1655
img6	GF	3.8828	7.1879	5.4636	45.7428	0.4640	6.2097
	RGF	5.5071	7.2264	8.0458	45.5467	0.5000	10.2460
	CNN	4.8318	7.1152	6.8625	51.0002	0.5378	9.1301
	CSMCA	5.2578	7.2167	7.0550	46.5250	0.6233	11.9430
	BRG	3.9936	7.4221	5.6747	50.3374	0.3925	7.1322
	PIAF	3.7770	7.5917	5.3198	52.3620	0.2461	5.7678
	本文方法	5.5097	7.6339	8.2027	54.6256	0.4846	9.5102
注：加粗数据表示最优结果

下载: 导出CSV

参考文献(21)

[1]	沈瑜, 陈小朋. 基于DLatLRR与VGG Net的红外与可见光图像融合[J]. 北京航空航天大学学报, 2021, 47(6): 1105-1114. SHEN Y, CHEN X P. Infrared and visible image fusion based on latent low-rank representation decomposition and VGG Net[J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(6): 1105-1114(in Chinese).
[2]	LI G F, LIN Y J, QU X D. An infrared and visible image fusion method based on multi-scale transformation and norm optimization[J]. Information Fusion, 2021, 71: 109-129. doi: 10.1016/j.inffus.2021.02.008
[3]	朱雯青, 汤心溢, 张瑞, 等. 基于边缘保持和注意力生成对抗网络的红外与可见光图像融合[J]. 红外与毫米波学报, 2021, 40(5): 696-708. ZHU W Q, TANG X Y, ZHANG R, et al. Infrared and visible image fusion based on edge-preserving and attention generative adversarial network[J]. Journal of Infrared and Millimeter Waves, 2021, 40(5): 696-708(in Chinese).
[4]	刘佳妮, 金伟其, 李力, 等. 自适应参考图像的可见光与热红外彩色图像融合算法[J]. 光谱学与光谱分析, 2016, 36(12): 3907-3914. LIU J N, JIN W Q, LI L, et al. Visible and infrared thermal image fusion algorithm based on self-adaptive reference image[J]. Spectroscopy and Spectral Analysis, 2016, 36(12): 3907-3914(in Chinese).
[5]	LUO Y Y, HE K J, XU D, et al. Infrared and visible image fusion based on visibility enhancement and hybrid multiscale decomposition[J]. Optik, 2022, 258: 168914. doi: 10.1016/j.ijleo.2022.168914
[6]	GUO Z Y, YU X T, DU Q L. Infrared and visible image fusion based on saliency and fast guided filtering[J]. Infrared Physics & Technology, 2022, 123: 104178.
[7]	MA J Y, ZHOU Y. Infrared and visible image fusion via gradientlet filter[J]. Computer Vision and Image Understanding, 2020, 197-198: 103016. doi: 10.1016/j.cviu.2020.103016
[8]	王满利, 王晓龙, 张长森. 基于动态范围压缩增强和NSST的红外与可见光图像融合算法[J]. 光子学报, 2022, 51(9): 277-291. WANG M L, WANG X L, ZHANG C S. Infrared and visible image fusion algorithm based on dynamic range compression enhancement and NSST[J]. Acta Photonica Sinica, 2022, 51(9): 277-291(in Chinese).
[9]	ZHOU Z Q, WANG B, LI S, et al. Perceptual fusion of infrared and visible images through a hybrid multi-scale decomposition with Gaussian and bilateral filters[J]. Information Fusion, 2016, 30: 15-26. doi: 10.1016/j.inffus.2015.11.003
[10]	LI X S, ZHOU F Q, TAN H S, et al. Multimodal medical image fusion based on joint bilateral filter and local gradient energy[J]. Information Sciences, 2021, 569: 302-325. doi: 10.1016/j.ins.2021.04.052
[11]	HU J W, LI S T. The multiscale directional bilateral filter and its application to multisensor image fusion[J]. Information Fusion, 2012, 13(3): 196-206. doi: 10.1016/j.inffus.2011.01.002
[12]	周志强, 汪渤, 李立广, 等. 基于双边与高斯滤波混合分解的图像融合方法[J]. 系统工程与电子技术, 2016, 38(1): 8-13. ZHOU Z Q, WANG B, LI L G, et al. Image fusion based on a hybrid decomposition via bilateral and Gaussian filters[J]. Systems Engineering and Electronics, 2016, 38(1): 8-13(in Chinese).
[13]	刘峰, 沈同圣, 马新星. 交叉双边滤波和视觉权重信息的图像融合[J]. 仪器仪表学报, 2017, 38(4): 1005-1013. doi: 10.3969/j.issn.0254-3087.2017.04.027 LIU F, SHEN T S, MA X X. Image fusion via cross bilateral filter and visual weight information[J]. Chinese Journal of Scientific Instrument, 2017, 38(4): 1005-1013(in Chinese). doi: 10.3969/j.issn.0254-3087.2017.04.027
[14]	戴进墩, 刘亚东, 毛先胤, 等. 基于FDST和双通道PCNN的红外与可见光图像融合[J]. 红外与激光工程, 2019, 48(2): 67-74. DAI J D, LIU Y D, MAO X Y, et al. Infrared and visible image fusion based on FDST and dual-channel PCNN[J]. Infrared and Laser Engineering, 2019, 48(2): 67-74(in Chinese).
[15]	MA J L, ZHOU Z Q, WANG B, et al. Infrared and visible image fusion based on visual saliency map and weighted least square optimization[J]. Infrared Physics & Technology, 2017, 82: 8-17.
[16]	LIU Y, CHEN X, CHENG J A, et al. Infrared and visible image fusion with convolutional neural networks[J]. International Journal of Wavelets, Multiresolution and Information Processing, 2018, 16(3): 1850018. doi: 10.1142/S0219691318500182
[17]	LIU Y, CHEN X, WARD R K, et al. Medical image fusion via convolutional sparsity based morphological component analysis[J]. IEEE Signal Processing Letters, 2019, 26(3): 485-489. doi: 10.1109/LSP.2019.2895749
[18]	ZHANG Y, ZHANG L J, BAI X Z, et al. Infrared and visual image fusion through infrared feature extraction and visual information preservation[J]. Infrared Physics & Technology, 2017, 83: 227-237.
[19]	TANG L F, YUAN J T, ZHANG H, et al. PIAFusion: A progressive infrared and visible image fusion network based on illumination aware[J]. Information Fusion, 2022, 83-84: 79-92. doi: 10.1016/j.inffus.2022.03.007
[20]	张小利, 李雄飞, 李军. 融合图像质量评价指标的相关性分析及性能评估[J]. 自动化学报, 2014, 40(2): 306-315. ZHANG X L, LI X F, LI J. Validation and correlation analysis of metrics for evaluating performance of image fusion[J]. Acta Automatica Sinica, 2014, 40(2): 306-315(in Chinese).
[21]	王晓文, 赵宗贵, 汤磊. 一种新的红外与可见光图像融合评价方法[J]. 系统工程与电子技术, 2012, 34(5): 871-875. WANG X W, ZHAO Z G, TANG L. Novel quality metric for infrared and visible image fusion[J]. Systems Engineering and Electronics, 2012, 34(5): 871-875(in Chinese).