Fusion of Infrared Intensity and Polarized Images Based on Structure and Decomposition
-
摘要: 在一些特定环境下,红外传感器无法探测到目标时,需要将偏振技术与红外技术相融合。为了获得更清楚的融合图像,采用一种基于多尺度结构分解的图像融合方法实现红外光强与偏振图像融合。该算法提出将红外图像与偏振图分解成3个独立部分:平均强度、信号强度和信号结构。其中平均强度部分,采用一种反正切的权重函数进行融合,信号强度采用最大值的融合原则,而信号结构采用一种基于信号强度幂函数的加权平均方进行融合,最后重构得到融合图像。为了更快进行融合、降低计算的复杂度,将分解过程通过均值滤波代替,再通过上采样与下采样得到最终的融合图像。为了得到更好的融合图像,通过不同融合参数实验对比,选择较优的融合参数。最后实验表明使用所提出的反正切权重函数与融合参数设置,在与传统的多尺度算法的比较中,4项评价指标取得优势,且主观上保留更多的纹理细节、提升对比度以及抑制伪影。Abstract: In specific environments, when an infrared sensor cannot detect a target, it is necessary to integrate polarization and infrared technologies. To obtain a clearer fused image, this study adopted a method based on a multiscale structure and feature image fusion to realize infrared and polarization image fusion. The algorithm decomposed the infrared image and polarization map into three independent parts: average intensity, signal intensity, and signal structure. An arctangent weight function was proposed for fusion in the average intensity part, the signal intensity adopted the maximum fusion principle, and the signal structure adopted a weighted average square based on the power function of the signal intensity for fusion, and finally, the fused image was reconstructed. To fuse faster and reduce computational complexity, the decomposition process was replaced with mean filtering, and the final fused image was obtained by upsampling and downsampling. To obtain a better fusion image, better fusion parameters were selected through an experimental comparison of different fusion parameters. Experiments showed that by using the proposed arctangent weight function and fusion parameter setting, the four evaluation indexes had advantages over the traditional multiscale algorithm and subjectively retained more texture details, improved contrast, and suppressed artifacts.
-
表 1 5组融合图像下不同λ的平均质量评价
Table 1. Average quality evaluation of different lambda under 5 groups of fused images
Parameter lambda =5 lambda =10 lambda =30 lambda =60 lambda = 100 lambda = 200 EN 7.2050 7.1933 7.1731 7.1663 7.1633 7.1621 0.5614 0.5620 0.5615 0.5609 0.5608 0.5607 SCD 1.5584 1.5400 1.5492 1.5637 1.5718 1.5785 SD 9.6199 9.5843 9.6108 9.6562 9.6403 9.6628 VIF 0.7385 0.7375 0.7364 0.7360 0.7359 0.7357 MS_SSIM 0.9661 0.9653 0.9644 0.9642 0.9643 0.9643 SSIM 0.9695 0.9691 0.9684 0.9682 0.9681 0.9680 表 2 融合图像的客观评价指标
Table 2. Objective evaluation indexes of fused images
Image Evaluation CP DWT GP LP PCA RP SIDWT Proposed Airport EN 7.4038 6.5629 6.5538 6.5413 6.3577 7.2649 6.5136 6.4584 QAB/F 0.2645 0.4154 0.4438 0.4599 0.3590 0.3114 0.4735 0.5238 SCD 0.8556 1.1152 1.1035 1.3192 0.9929 1.3989 1.2233 0.7526 SD 8.9371 8.4546 8.7057 8.3978 9.1176 9.4301 8.4619 8.1108 VIF 0.7788 0.4912 0.5371 0.5760 0.5500 0.5795 0.5728 0.6035 MS_SSIM 0.5425 0.9414 0.9518 0.9685 0.9289 0.6326 0.9752 0.9713 SSIM 0.5021 0.9374 0.9541 0.9565 0.9217 0.5491 0.9715 0.9715 Road EN 7.4540 7.2851 7.1242 7.4016 7.4545 7.4713 7.2958 7.7729 QAB/F 0.2534 0.4296 0.4426 0.4726 0.4537 0.2949 0.4729 0.5103 SCD 1.4472 1.6042 1.5539 1.6783 0.3136 1.5853 1.6133 1.7381 SD 9.5398 10.4314 10.3399 10.3981 10.4773 9.9832 10.4107 10.4351 VIF 0.9118 0.5729 0.6390 0.6850 0.9295 0.5813 0.6657 0.8105 MS_SSIM 0.5506 0.8840 0.9042 0.9400 0.7670 0.6510 0.9382 0.9617 SSIM 0.5906 0.9179 0.9387 0.9495 0.6648 0.6875 0.9590 0.9575 Car EN 6.8170 6.9337 6.8247 6.9915 7.2436 6.9526 6.9374 7.4851 QAB/F 0.3286 0.5521 0.5660 0.6039 0.6784 0.2854 0.6042 0.6583 SCD 1.3841 1.4742 1.4815 1.5420 0.4592 1.4293 1.4974 1.7339 SD 9.2526 9.6851 9.6849 9.8760 9.8916 9.6622 9.8453 9.7009 VIF 0.5617 0.6112 0.6733 0.7692 1.0708 0.4070 0.6982 0.9140 MS_SSIM 0.6908 0.8840 0.9042 0.9400 0.7670 0.6510 0.9382 0.9817 SSIM 0.6762 0.9395 0.9444 0.9630 0.8894 0.7033 0.9688 0.9753 Windows EN 7.3378 6.6612 6.5249 7.4594 7.2594 7.1577 6.6466 7.2811 QAB/F 0.1769 0.4799 0.4937 0.3786 0.3686 0.2412 0.5176 0.5272 SCD 1.1893 1.6507 1.6235 0.3339 0.3139 1.4610 1.6734 1.8535 SD 9.1006 9.3828 9.3866 10.9746 10.9446 8.8531 9.3979 10.7113 VIF 0.7658 0.3381 0.3302 1.0025 1.0125 0.4664 0.3704 0.4488 MS_SSIM 0.3876 0.9059 0.9236 0.6470 0.6370 0.4996 0.9550 0.9461 SSIM 0.4673 0.9554 0.9644 0.6249 0.6148 0.5762 0.9782 0.9688 Outdoor EN 6.7270 6.6479 6.4716 6.7282 6.6535 6.6863 6.6336 7.0276 QAB/F 0.5419 0.5050 0.5196 0.5485 0.6341 0.4022 0.5335 0.5873 SCD 1.5299 1.5146 1.4948 1.6165 0.0819 1.6062 1.5021 1.7136 SD 8.2680 8.5149 8.2519 8.4752 7.0202 8.1989 8.4533 9.1415 VIF 0.7738 0.5876 0.6941 0.7670 0.9916 0.5369 0.7160 0.9157 MS_SSIM 0.8585 0.9122 0.9260 0.9607 0.6508 0.8497 0.9586 0.9700 SSIM 0.8927 0.9452 0.9555 0.9689 0.8242 0.8731 0.9754 0.9744 Average values of 5 fused images EN 7.1479 6.8182 6.6998 7.0244 6.9937 7.1066 6.8054 7.2050 QAB/F 0.3131 0.4764 0.4931 0.4927 0.4988 0.3070 0.5204 0.5614 SCD 1.2812 1.4718 1.4514 1.2980 0.4323 1.4962 1.5019 1.5584 SD 9.0196 9.2938 9.2738 9.6243 9.4903 9.2255 9.3138 9.6199 VIF 0.7584 0.5202 0.5748 0.7599 0.9109 0.5142 0.6046 0.7385 MS_SSIM 0.6060 0.9139 0.9252 0.8967 0.7762 0.6632 0.9573 0.9661 SSIM 0.6258 0.9391 0.9514 0.8925 0.7830 0.6778 0.9706 0.9695 -
[1] 张肃, 付强, 段锦, 等. 基于提升小波的低对比度目标偏振识别技术[J]. 光学学报, 2015, 35(2): 0211002. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201502017.htmZHANG S, FU Q, DUAN J, et al. Low contrast target polarization recognition technology based on lifting wavelet[J]. Acta Optica Sinica, 2015, 35(2): 0211002. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201502017.htm [2] 陈潮起, 孟祥超, 邵枫, 等. 一种基于多尺度低秩分解的红外与可见光图像融合方法[J]. 光学学报, 2020, 40(11): 1110001. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB202011008.htmCHEN C, MENG X, SHAO F, et al. Infrared and visible image fusion method based on multiscale low-rank decomposition[J]. Acta Optica Sinica, 2020, 40(11): 1110001. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB202011008.htm [3] CHEN J, LI X J, LUO L B, et al. Infrared and visible image fusion based on target-enhanced multiscale transform decomposition[J]. Information Sciences, 2020, 508: 64-78. doi: 10.1016/j.ins.2019.08.066 [4] 汪美玉, 陈代梅, 赵根保. 基于目标提取与拉普拉斯变换的红外和可见光图像融合算法[J]. 激光与电子学进展, 2017, 54(1): 011002. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201701013.htmWANG M, CHEN D, ZHAO G. Image fusion algorithm of infrared and visible images based on target extraction and laplace transformation[J]. Laser & Optoelectronics Progress, 2017, 54(1): 011002. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ201701013.htm [5] 杨九章, 刘炜剑, 程阳. 基于对比度金字塔与双边滤波的非对称红外与可见光图像融合[J]. 红外技术, 2021, 43(9): 840-844. http://hwjs.nvir.cn/article/id/1c7de46d-f30d-48dc-8841-9e8bf3c91107YANG J, LIU W, CHEN Y. Asymmetric infrared and visible image fusion based on contrast pyramid and bilateral filtering[J]. Infrared Technology, 2021, 43(9): 840-844. http://hwjs.nvir.cn/article/id/1c7de46d-f30d-48dc-8841-9e8bf3c91107 [6] 黄光华, 倪国强, 张彬. 一种基于视觉阈值特性的图像融合方法[J]. 北京理工大学学报, 2006(10): 907-911. doi: 10.3969/j.issn.1001-0645.2006.10.015HUANG G, NI G, ZHANG B. Image fusion by a visual threshold based pyramid[J]. Transactions of Beijing Institute of Technology, 2006(10): 907-911. doi: 10.3969/j.issn.1001-0645.2006.10.015 [7] 李建林, 俞建成, 孙胜利. 基于梯度金字塔图像融合的研究[J]. 科学技术与工程, 2007(22): 5818-5822. https://www.cnki.com.cn/Article/CJFDTOTAL-KXJS200722017.htmLI J, YU J, SUN S. Study of image fusion based on grad pyramid algorithm[J]. Science Technology and Engineering, 2007(22): 5818-5822. https://www.cnki.com.cn/Article/CJFDTOTAL-KXJS200722017.htm [8] 杨风暴, 董安冉, 张雷, 等. DWT、NSCT和改进PCA协同组合红外偏振图像融合[J]. 红外技术, 2017, 39(3): 201-208. http://hwjs.nvir.cn/article/id/hwjs201703001YANG F, DONG A, ZHANG L, et al. Infrared polarization image fusion using the synergistic combination of DWT, NSCT and improved PCA[J]. Infrared Technology, 2017, 39(3): 201-208. http://hwjs.nvir.cn/article/id/hwjs201703001 [9] 安富, 杨风暴, 李伟伟, 等. 基于DWT的红外偏振与光强图像的融合[J]. 光电技术应用, 2013, 28(2): 18-22. https://www.cnki.com.cn/Article/CJFDTOTAL-GDYG201302008.htmAN F, YANG F, LI W, et al. Fusion of infrared polarization and intensity images based on DWT[J]. Electro-Optic Technology Application, 2013, 28(2): 18-22. https://www.cnki.com.cn/Article/CJFDTOTAL-GDYG201302008.htm [10] WANG X, WEI Y L, LIU Fu. A new multi-source image sequence fusion algorithm based on SIDWT[C]//2013 Seventh International Conference on Image and Graphics, 2013: 568-571. [11] YANG F, WEI H. Fusion of infrared polarization and intensity images using support value transform and fuzzy combination rules[J]. Infrared Physics & Technology, 2013, 60: 235-243. [12] ZHU Pan, HUANG Zhanhua. A fusion method for infrared-visible image and infrared-polarization image based on multi-scale center-surround top-hat transform[J]. Optical Review, 2017, 24(3): 1-13. [13] LI H, QI X B, XIE W Y. Fast infrared and visible image fusion with structural decomposition[J]. Knowledge-Based Systems, 2020, 204: 106182. [14] MA K, HUI L, YONG H, et al. Robust multi-exposure image fusion: a structural patch decomposition approach[J]. IEEE Transactions on Image Processing, 2017, 26(5): 2519-2532. [15] KOU F, LI Z G, WEN C Y, et al. Multi-scale exposure fusion via gradient domain guided image filtering[C]//2017 IEEE International Conference on Multimedia and Expo, 2017: 1105-1110. [16] LI H, JIA X X, ZHANG Lei. Clustering based content and color adaptive tone mapping[J]. Computer Vision and Image Understanding, 2018, 168: 37-49. [17] Aardt V Jan. Assessment of image fusion procedures using entropy, image quality, and multispectral classification[J]. Journal of Applied Remote Sensing, 2008, 2(1): 1-28. [18] Piella G, Heijmans H. A new quality metric for image fusion[C]// International Conference on Image Processing, 2003: 173-176. [19] Aslantas V, Bende E. A new image quality metric for image fusion: The sum of the correlations of differences[J]. AEU-International Journal of Electronics and Communications, 2015, 69(12): 1890-1896. [20] Altman D G, Bland J M. Statistics notes - Standard deviations and standard errors[J]. British Medical Journal, 2005, 331(7521): 903-903. [21] Sheikh H R, Bovik A C. Image information and visual quality[J]. IEEE Transaction on Image Processing, 2006, 15(2): 430-444. [22] WANG Z, Simoncelli E P, Bovik A C. Multiscale structural similarity for image quality assessment[C]//Proc IEEE Asilomar Conference on Signals, 2003: 1398-1402. [23] ZHOU W, Bovik A C, Sheikh H R, et al. Image quality assessment: from error visibility to structural similarity[J]. IEEE Trans Image Process, 2004, 13(4): 600-612.