基于改进小波阈值函数和全尺度Retinex的红外图像融合增强算法

童耀南; 杨海涛; 曹志奇; 崔建山; 刘智

基于改进小波阈值函数和全尺度Retinex的红外图像融合增强算法

湖南理工学院信息科学与工程学院复杂网络与嵌入式系统湖南省高校重点实验室, 湖南岳阳 414006

基金项目:

湖南省重点实验室 2019TP1014

湖南省教育厅重点科研项目 21A0403

湖南省教育厅重点科研项目 21A0405

湖南省研究生科研创新项目 QL20210255

详细信息

作者简介:
童耀南（1977-），男，博士，教授，湖南平江人，主要从事物联网技术、嵌入式人工智能、模数混合实时信号处理。E-mail: tongyaon@hnist.edu.cn

中图分类号: TP391.41
计量
- 文章访问数: 117
- HTML全文浏览量: 25
- PDF下载量: 41
出版历程
- 收稿日期: 2023-05-15
- 修回日期: 2023-09-05
- 刊出日期: 2024-03-19

Infrared Image Fusion Enhancement Algorithm Based on Improved Wavelet Threshold Function and Full-Scale Retinex

College of Information Science and Engineering, Hunan Institute of Science and Technology & Key Laboratory of Complex Network and Embedded System, College of Hunan Province, Yueyang 414006, China

摘要

摘要: 针对现有红外图像增强算法存在信噪比低、细节模糊、清晰度差等问题，本文提出基于改进小波阈值函数和全尺度Retinex的红外图像融合增强算法。首先，为克服尺度参数固定和光线散射导致红外图像退化的问题，利用大气透射率得到Retinex尺度参数的全尺度映射图，从而有效提高图像的清晰度，并将输入图像和使用全尺度Retinex处理后的输入图像作为算法的第一个输入和第二个输入。其次，为解决传统小波阈值函数在图像降噪过程中存在伪影、细节丢失等问题，设计改进小波阈值函数，通过引入尺度因子，在计算每层高频子图小波系数后，能根据该层数自适应调整尺度因子，并引入调节因子，结合指数函数，使该函数不仅能抑制高频子图噪声，还能极大程度保留细节信息。然后，使用小波图像融合的方式融合输入的高频子图和低频子图，进一步提高输出图像的纹理细节。主客观仿真结果表明，所提算法比其它对比算法具有更好的降噪和细节突出能力，并能提高红外图像的人眼视觉效果。最后，本文算法应用于红外成像模块采集的红外图像增强，效果良好，表明本文方法具有实用性。
- 改进小波阈值函数 /
- 全尺度Retinex /
- 红外图像增强 /
- 图像融合
Abstract: This paper proposes an infrared image fusion enhancement algorithm based on an improved wavelet threshold function and full-scale Retinex to address the problems of low signal-to-noise ratio, fuzzy detail, and poor clarity in existing infrared image enhancement algorithms. First, to overcome the degradation of infrared images caused by fixed-scale parameters and light scattering, a full-scale map of Retinex-scale parameters was obtained using atmospheric transmittance to improve image clarity. The input image and processed image with full-scale Retinex were used as the first and second inputs of the algorithm, respectively. Second, an improved wavelet threshold function was designed to solve the problems of artifacts and detail loss in the image-denoising process of the traditional wavelet threshold function. The threshold function introduces a scaling factor that can be adjusted adaptively according to the number of layers after calculating the wavelet coefficient of the high-frequency subgraph of each layer. An adjustment factor was introduced and combined with an exponential function to suppress the high-frequency subgraph noise and preserve detailed information. The high- and low-frequency subgraphs of the above two inputs were then fused using wavelet image fusion to improve the texture details of the output images. The simulation results demonstrate that the proposed algorithm outperforms other comparison algorithms regarding noise reduction and detail highlighting capabilities, enhancing the visual quality of infrared images for the human eye. Finally, this algorithm was applied to enhance infrared images collected by an infrared imaging module, and the experimental results showed that the proposed method is practical.
- improved wavelet threshold function /
- full-scale Retinex /
- infrared image enhancement /
- image fusion

HTML全文

图 1 本文算法流程图

Figure 1. Algorithm flow chart of this paper

下载: 全尺寸图片幻灯片

图 2 不同阈值下的小波阈值函数

Figure 2. Wavelet threshold function under different threshold values

下载: 全尺寸图片幻灯片

图 3 PSNR和SSIM的变化曲线

Figure 3. Variation curves of PSNR and SSIM

下载: 全尺寸图片幻灯片

图 4 PSNR随分解层数变化曲线

Figure 4. PSNR curves with decomposition layers

下载: 全尺寸图片幻灯片

图 5 尺度映射

Figure 5. Scale mapping

下载: 全尺寸图片幻灯片

图 6 全尺度Retinex算法流程图

Figure 6. Full scale Retinex algorithm flow chart

下载: 全尺寸图片幻灯片

图 7 场景1红外图像增强算法对比图

Figure 7. Comparison of infrared image enhancement algorithms in scene 1

下载: 全尺寸图片幻灯片

图 8 场景2红外图像增强算法对比图

Figure 8. Comparison of infrared image enhancement algorithms in scene 2

下载: 全尺寸图片幻灯片

图 9 场景3红外图像增强算法对比图

Figure 9. Comparison of infrared image enhancement algorithms in scene 3

下载: 全尺寸图片幻灯片

图 10 场景4红外图像增强算法对比图

Figure 10. Comparison of infrared image enhancement algorithms in scene 4

下载: 全尺寸图片幻灯片

图 11 不同降噪方法的测试结果

Figure 11. Test results of different noise reduction methods

下载: 全尺寸图片幻灯片

图 12 搭建的红外图像增强上位机流程

Figure 12. Infrared image enhancement host computer process

下载: 全尺寸图片幻灯片

图 13 红外成像模块及主观评价指标结果

Figure 13. Infrared imaging module and subjective evaluation index results

下载: 全尺寸图片幻灯片

表 1 场景1、2、3、4的客观评价指标结果

Table 1 Results of objective evaluation indicators in scenarios 1, 2, 3 and 4

Image	Evaluation indicators	f	BF & DRP	GIF	MSR	DCP	Document[18]	Document[19]	Proposed in this paper
Scenarios 1	RMSC	28.68	27.76	31.21	29.58	29.44	32.63	14.70	33.95
	PSNR	-	16.05	23.23	22.66	18.50	23.96	31.84	30.92
	DE	6.75	6.67	7.03	6.76	6.72	6.93	6.78	7.05
	SSIM	-	0.91	0.84	0.82	0.88	0.85	0.82	0.97
Scenarios 2	RMSC	15.90	16.72	29.06	18.36	30.92	32.99	17.58	32.91
	PSNR	-	15.49	18.70	13.88	16.60	22.21	17.29	24.42
	DE	6.02	6.07	7.15	7.09	6.87	6.69	6.13	7.16
	SSIM	-	0.89	0.83	0.81	0.77	0.82	0.78	0.96
Scenarios 3	RMSC	24.48	25.33	27.02	22.18	23.15	33.73	31.92	32.68
	PSNR	-	16.66	25.72	20.25	23.55	24.36	24.66	25.44
	DE	6.61	6.63	6.90	6.71	6.43	6.93	6.85	7.02
	SSIM	-	0.81	0.90	0.81	0.90	0.86	0.87	0.96
Scenarios 4	RMSC	44.75	45.46	47.29	45.13	42.69	42.77	46.29	49.81
	PSNR	-	22.23	21.59	15.10	23.38	16.51	16.80	24.71
	DE	7.05	7.10	7.23	7.12	7.07	7.21	7.16	7.25
	SSIM	-	0.83	0.91	0.87	0.93	0.76	0.88	0.94

下载: 导出CSV

表 2 不同阈值降噪方法的客观评价值

Table 2 Objective evaluation value of noise reduction methods with different threshold values

Noise variance	Evaluation indicators	Hard threshold function	Soft threshold function	Improved wavelet threshold function
0.01	PSNR	27.76	27.66	28.62
0.01	SSIM	0.92	0.92	0.93
0.02	PSNR	25.37	25.33	25.83
0.02	SSIM	0.87	0.86	0.88
0.03	PSNR	23.96	23.89	24.24
0.03	SSIM	0.82	0.82	0.83

下载: 导出CSV

表 3 本文算法在场景5和场景6的客观评价结果

Table 3 Objective evaluation results of the algorithm in scenarios 5 and scenarios 6

Evaluation indicators	Scenario 5	Scenario 6
RMSC	27.51	30.51
PSNR	19.86	21.39
DE	7.66	7.23
SSIM	0.51	0.63

下载: 导出CSV

参考文献(19)

[1]	Gowen A A, Tiwari B K, Cullen P J, et al. Applications of thermal imaging in food quality and safety assessment[J]. Trends in Food Science & Technology, 2010, 21(4): 190-200.
[2]	赵明珠, 张艳, 朱应燕. 基于红外热成像的早期疾病检测技术的研究进展[J]. 激光与光电子学进展, 2021, 58(8): 28-38. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ202108003.htm ZHAO Mingzhu, ZHANG Yan, ZHU Yingyan. Research progress of early disease detection technology based on infrared thermography[J]. Advances in Lasers and Optoelectronics, 2021, 58(8): 28-38. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ202108003.htm
[3]	YUAN C, LIU Z, ZHANG Y. Fire detection using infrared images for UAV-based forest fire surveillance[C]//International Conference on Unmanned Aircraft Systems (ICUAS), 2017: 567-572.
[4]	WANG L, Leedham G, Cho S Y. Infrared imaging of hand vein patterns for biometric purposes[J]. IET Computer Vision, 2007, 1(3): 113-122. DOI: 10.1049/iet-cvi:20070009
[5]	Taheri Garavand A, Ahmadi H, Omid M, et al. An intelligent approach for cooling radiator fault diagnosis based on infrared thermal image processing technique[J]. Applied Thermal Engineering, 2015, 87: 434-443. DOI: 10.1016/j.applthermaleng.2015.05.038
[6]	郝建新. 基于小波变换与Retinex的电路板红外图像增强技术[J]. 红外技术, 2015, 37(12): 1036-1040. http://hwjs.nvir.cn/article/id/hwjs201512009 HAO Jianxin. Circuit board infrared image enhancement based on wavelet transform and Retinex[J]. Infrared Technology, 2015, 37(12): 1036-1040. http://hwjs.nvir.cn/article/id/hwjs201512009
[7]	尹士畅, 喻松林. 基于小波变换和直方图均衡的红外图像增强[J]. 激光与红外, 2013, 43(2): 225-228. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201302025.htm YIN Shichang, YU Songlin. Infrared image enhancement based on wavelet transform and histogram equalization[J]. Laser and Infrared, 2013, 43(2): 225-228. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201302025.htm
[8]	CAI B, XU X, GUO K, et al. A joint intrinsic-extrinsic prior model for Retinex[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 4000-4009.
[9]	ZHANG X, WANG J. Contourlet-based non-local mean via Retinex theory for robot infrared image enhancement[J]. EAI Endorsed Transactions on Scalable Information Systems, 2022, 9(4): e10-e10.
[10]	王晨, 汤心溢, 高思莉. 基于人眼视觉的红外图像增强算法研究[J]. 激光与红外, 2017, 47(1): 114-118. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201701023.htm WANG Chen, TANG Xinyi, GAO Sili. Research on infrared image enhancement algorithm based on human eye vision[J]. Laser and Infrared, 2017, 47(1): 114-118. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201701023.htm
[11]	龚昌来, 罗聪, 杨冬涛, 等. 一种基于平稳小波域的红外图像增强方法[J]. 激光与红外, 2013, 43(6): 703-707. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201306024.htm GONG Changlai, LUO Cong, YANG Dongtao, et al. An infrared image enhancement method based on smooth wavelet domain[J]. Laser and Infrared, 2013, 43(6): 703-707. https://www.cnki.com.cn/Article/CJFDTOTAL-JGHW201306024.htm
[12]	Donoho D L, Johnstone I M. Ideal spatial adaptation by wavelet shrinkage[J]. Biometrika, 1994, 81(3): 425-455. DOI: 10.1093/biomet/81.3.425
[13]	黄玉昌, 侯德文. 基于改进小波阈值函数的指纹图像去噪[J]. 计算机工程与应用, 2014, 50(6): 179-181, 209. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201406039.htm HUANG Yuchang, HOU Dewen. Fingerprint image denoising based on improved wavelet threshold function[J]. Computer Engineering and Applications, 2014, 50(6): 179-181, 209. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201406039.htm
[14]	DU J, LI W, XIAO B, et al. Union Laplacian pyramid with multiple features for medical image fusion[J]. Neurocomputing, 2016, 194: 326-339. DOI: 10.1016/j.neucom.2016.02.047
[15]	刘志强, 朱大奇. 一种新型小波图像融合的水下目标增强算法[J]. 控制工程, 2022, 29(12): 2235-2243. https://www.cnki.com.cn/Article/CJFDTOTAL-JZDF202212008.htm LIU Zhiqiang, ZHU Daqi. A novel underwater target enhancement algorithm for wavelet image fusion[J]. Control Engineering, 2022, 29(12): 2235-2243. https://www.cnki.com.cn/Article/CJFDTOTAL-JZDF202212008.htm
[16]	易清明, 陈明敏, 石敏. 一种改进的小波去噪方法在红外图像中应用[J]. 计算机工程与应用, 2016, 52(1): 173-177. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201601033.htm YI Qingming, CHEN Mingmin, SHI Min. An improved wavelet denoising method in infrared images[J]. Computer Engineering and Applications, 2016, 52(1): 173-177. https://www.cnki.com.cn/Article/CJFDTOTAL-JSGG201601033.htm
[17]	徐兴贵, 杨平, 刘永利. 基于全尺度Retinex算法的夜间图像去雾[J]. 微电子学与计算机, 2017, 34(7): 132-136. https://www.cnki.com.cn/Article/CJFDTOTAL-WXYJ201707028.htm XU Xinggui, YANG Ping, LIU Yongli. Nighttime image defogging based on full-scale Retinex algorithm[J]. Microelectronics and Computers, 2017, 34(7): 132-136. https://www.cnki.com.cn/Article/CJFDTOTAL-WXYJ201707028.htm
[18]	占必超, 吴一全, 纪守新. 基于平稳小波变换和Retinex的红外图像增强方法[J]. 光学学报, 2010, 30(10): 2788-2793. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201010007.htm ZHAN Bichao, WU Yiquan, JI Shouxin. Infrared image enhancement method based on smooth wavelet transform and Retinex[J]. Journal of Optics, 2010, 30(10): 2788-2793. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201010007.htm
[19]	陈文艺, 杨承勋, 杨辉. 引导滤波和对数变换算法融合的多尺度Retinex红外图像增强[J]. 红外技术, 2022, 44(4): 397-403. http://hwjs.nvir.cn/article/id/f1fcd3be-4a81-4b25-ad02-5f8c035f0be2 CHEN Wenyi, YANG Chengxun, YANG Hui. Multiscale Retinex infrared image enhancement by fusion of bootstrap filtering and logarithmic transformation algorithms[J]. Infrared Technology, 2022, 44(4): 397-403. http://hwjs.nvir.cn/article/id/f1fcd3be-4a81-4b25-ad02-5f8c035f0be2

施引文献(6)

期刊类型引用(4)

1.	冉兴程，李广伟，雷永. 基于模糊贴近算法的海量异常激光数据挖掘方法研究. 自动化与仪器仪表. 2019(04): 142-145 . 百度学术
2.	王茜，平金珍，张永华. 星载光交换信道数据挖掘系统设计. 激光杂志. 2018(03): 151-154 . 百度学术
3.	杨鹏飞，廖秀英，程辉，帅慕容，谢贻文. 基于光谱梯度角与光谱信息散度评价指标的遥感影像融合方法比较. 测绘工程. 2018(11): 51-55 . 百度学术
4.	董熙，李存军，夏朝宗，王二丽，杨新兵，周静平，胡海棠. 北京平原造林地Pleiades卫星影像融合评价. 林业资源管理. 2016(03): 128-134 . 百度学术