-
摘要:
针对偏振光谱图像融合方法在地物混杂背景遥感探测中多尺度变换融合图像边缘轮廓细节模糊、对比度不佳的问题，采用一种基于非下采样轮廓波变换的稀疏表示与引导滤波器相结合的图像融合方法，以改善融合图像的质量和视觉效果。首先，该方法通过非下采样轮廓波变换对光谱图像和偏振图像进行多尺度多方向分解，进而将图像分解成不同子带内的特征信息。其次，低频子带采用稀疏表示融合，从而降低融合图像中物体对比度损失。此外，采用引导滤波器融合高频子带，达到增强图像轮廓细节信息的效果。最后，对低频与高频融合系数进行非下采样轮廓波逆变换最终得出融合图像。分析表明融合图像对比度相对于原始光谱图像与偏振度图像分别提升了54.5%和15.4%，更容易区分混杂背景下阴影中的物体。基于此方法对偏振光谱成像仪所采集的不同波长下的光谱与偏振图像进行融合，并实现真彩还原。真彩还原图像证明此融合方法在保留混杂背景下的环境信息的同时实现物体和背景区分，有效提高了偏振光谱遥感探测成像的图像质量，有助于提升偏振光谱遥感探测成像中图像信息的完整性和真实性，扩大其在复杂环境遥感探测和图像识别中的应用范围。
Abstract:To address the issues of blurred edge details and poor contrast in multi-scale transform fused images obtained using remote sensing detection methods for mixed background features, an image fusion approach that combines the sparse representation of non-downsampled contour wavelet transform and a guided filter was utilised to enhance the quality and visual appearance of the fused images. This method involved several steps: Firstly, a multi-scale and multi-directional decomposition was performed on both spectral and polarimetric images using non-downsampled contour wavelet transform to isolate the feature information in each subband; secondly, the low-frequency subbands were fused using a sparse representation approach to minimize the loss of contrast in the fused image; additionally, the high-frequency subbands were fused through a bootstrap filter to enhance the detail information and the contours of the image; finally, the low-frequency and high-frequency fusion coefficients were inverted using non-downsampled contour wavelet inversion to generate the final fused image. Analysis indicates that this method has increased the contrast of the fused image by up to 54.5% relative to the original spectral image and by 15.4% compared to the polarimetric image. This has resulted in a fused image in which it is easier to distinguish objects in shadows within a mixed background. This method was used to fuse spectral and polarimetric images captured by a polarimetric spectral imager at different wavelengths, which resulted in true-colour reproduction. These true-colour restored images demonstrate that this fusion method retains environmental information within the mixed background while distinguishing the object from the background, effectively improving the image quality of polarization spectral remote sensing detection imaging. This approach enhances the integrity and authenticity of image information in polarization spectral remote sensing detection imaging, thereby expanding its application scope in remote sensing detection of complex environments and image recognition.
-
图 2 基于NSCT的稀疏表示与引导滤波器的图像融合方法流程图
Figure 2. Flowchart of the fusion method discussed in this paper
图 9 四种波长下的不同融合方法的客观评价指标
Figure 9. Objective evaluation indexes of different fusion methods at four wavelengths
表 1 互信息MI与加权融合质量指数Qw两种客观评价指标评价结果
Table 1. Results of two objective evaluations of mutual information MI and weighted fusion quality index Qw
评价
指标融合规则 区域方差匹配度 取平均 权重融合 稀疏表示 MI 1.6685 1.4523 1.6238 1.9777 Qw 0.6408 0.6324 0.7241 0.7822 
下载: 导出CSV
表 2 结构信息相似度SSIM和边缘保持度Qab/f客观指标评价指标结果
Table 2. Evaluation metric results for structural information similarity SSIM and edge retention Qab/f objective metrics
评价
指标融合规则 PCNN 边缘探测 绝对值取大 引导滤波 SSIM 0.7354 0.6821 0.6378 0.7562 Qab/f 0.5916 0.4481 0.4507 0.5988
下载: 导出CSV
表 3 系统指标
Table 3. System indicators
性能指标 参数值 波长范围 400 nm−900 nm 光谱分辨率 优于2 nm F/# 3 视场角 ±4.1° 分辨率 2448*2048 像元尺寸 3.45 μm*3.45 μm 光学尺寸 2/3 inch 帧率 36 fps@2448*2048
下载: 导出CSV
表 4 参考方法与本文方法的各分量融合图像评价指标
Table 4. The evaluation indexes of image fusion for each component of the reference method and the proposed method
融合图像 评价指标 MI SSIM Qab/f Qw 参考方法I融合结果 2.7623 0.7182 0.6147 0.8367 本文I融合结果 2.8031 0.7387 0.6393 0.8625 参考方法DoLP融合结果 1.9765 0.6376 0.5733 0.7615 本文DoLP融合结果 2.7342 0.7023 0.6142 0.8222 参考方法AoP融合结果 1.6681 0.5679 0.4935 0.6138 本文AoP融合结果 1.7437 0.6755 0.5883 0.7004
下载: 导出CSV
表 5 参考方法与本文方法的HSI伪彩色图像评价指标
Table 5. The evaluation indexes of HSI pseudo-colored images for the reference method and the proposed method
融合图像 评价指标 PSNR SSIM 参考方法HSI融合结果 22.81dB 0.6431 本文方法HSI融合结果 28.52dB 0.6947
下载: 导出CSV
表 6 不同融合方法结果评价指标
Table 6. Result evaluation indexes of different fusion methods
融合方法 评价指标 MI SSIM Qab/f Qw CNN 2.2343 0.6434 0.5662 0.7787 GFF 1.7685 0.5862 0.5137 0.7507 LP-SR 2.1132 0.6361 0.5354 0.7427 CVT-SR 2.2356 0.6053 0.5833 0.7706 NSCT-SR 2.5835 0.5965 0.5350 0.7582 本文方法 2.7342 0.7023 0.6142 0.8222
下载: 导出CSV
表 7 PSNR\SSIM评价指标
Table 7. PSNR\SSIM evaluation index
评价
指标真彩图像 原始真彩图像 融合后的真彩图像 PSNR 33.72 dB 36.87 dB SSIM 0.8672 0.9104
下载: 导出CSV
-
[1] HE Z J, ZHANG Z Z, GUO M Q, et al. Adaptive unsupervised-shadow-detection approach for remote-sensing image based on multichannel features[J]. Remote Sensing, 2022, 14(12): 2756. doi: 10.3390/rs14122756 [2] LI J X, HONG D F, GAO L R, et al. Deep learning in multimodal remote sensing data fusion: a comprehensive review[J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 112: 102926. doi: 10.1016/j.jag.2022.102926 [3] 张璐, 李博, 李寒霜, 等. 超光谱分辨率紫外双通道共光路成像光谱仪设计[J]. 中国光学(中英文),2022,15(5):1029-1037. doi: 10.37188/CO.2022-0125ZHANG L, LI B, LI H SH, et al. Hyperspectral resolution ultraviolet dual channel common optical path imaging spectrometer[J]. Chinese Optics, 2022, 15(5): 1029-1037. (in Chinese). doi: 10.37188/CO.2022-0125 [4] 周俊焯, 郝佳, 余晓畅, 等. 面向偏振成像的超构表面研究进展[J]. 中国光学(中英文),2023,16(5):973-995. doi: 10.37188/CO.2022-0234ZHOU J ZH, HAO J, YU X CH, et al. Recent advances in metasurfaces for polarization imaging[J]. Chinese Optics, 2023, 16(5): 973-995. (in Chinese). doi: 10.37188/CO.2022-0234 [5] WANG J Y, SHI H D, LIU J N, et al. Compressive space-dimensional dual-coded hyperspectral polarimeter (CSDHP) and interactive design method[J]. Optics Express, 2023, 31(6): 9886-9903. (查阅网上资料, 未能确认标黄作者信息, 请确认) .WANG J Y, SHI H D, LIU J N, et al. . Compressive space-dimensional dual-coded hyperspectral polarimeter (CSDHP) and interactive design method[J]. Optics Express, 2023, 31(6): 9886-9903. (查阅网上资料, 未能确认标黄作者信息, 请确认). [6] MA J Y, MA Y, LI CH. Infrared and visible image fusion methods and applications: a survey[J]. Information Fusion, 2019, 45: 153-178. doi: 10.1016/j.inffus.2018.02.004 [7] KAUR H, KOUNDAL D, KADYAN V. Image fusion techniques: a survey[J]. Archives of Computational Methods in Engineering, 2021, 28(7): 4425-4447. doi: 10.1007/s11831-021-09540-7 [8] ZHAO Y, ZHANG L, ZHANG D, et al. Object separation by polarimetric and spectral imagery fusion[J]. Computer Vision and Image Understanding, 2009, 113(8): 855-866. doi: 10.1016/j.cviu.2009.03.002 [9] 于洵, 杨烨, 姜旭, 等. 基于偏振光谱成像的目标识别方法研究[J]. 应用光学,2016,37(4):537-541. doi: 10.5768/JAO201637.0402001YU X, YANG Y, JIANG X, et al. Recognition of camouflage targets by polarization spectral imaging system[J]. Journal of Applied Optics, 2016, 37(4): 537-541. (in Chinese). doi: 10.5768/JAO201637.0402001 [10] 牛思聪. 基于生成对抗网络的偏振高光谱图像处理与分类算法研究[D]. 哈尔滨: 哈尔滨工业大学, 2021.NIU S C. Research on polarization hyperspectral image processing and classification algorithm based on generative countermeasure network[D]. Harbin: Harbin Institute of Technology, 2021. (in Chinese). [11] 钟菁菁, 刘骁, 王雪霁, 等. 偏振光谱多维信息的重构融合算法[J]. 光谱学与光谱分析,2023,43(4):1254-1261.ZHONG J J, LIU X, WANG X J, et al. A multidimensional information fusion algorithm for polarization spectrum reconstruction based on nonsubsampled contourlet transform[J]. Spectroscopy and Spectral Analysis, 2023, 43(4): 1254-1261. (in Chinese). [12] LIU Y, YAN B Y, ZHANG R Z, et al. Multi-scale mixed attention network for CT and MRI image fusion[J]. Entropy, 2022, 24(6): 843. doi: 10.3390/e24060843 [13] 朱攀, 刘泽阳, 黄战华. 基于DTCWT和稀疏表示的红外偏振与光强图像融合[J]. 光子学报,2017,46(12):1210002. doi: 10.3788/gzxb20174612.1210002ZHU P, LIU Z Y, HUANG ZH H. Infrared polarization and intensity image fusion based on dual-tree complex wavelet transform and sparse representation[J]. Acta Photonica Sinica, 2017, 46(12): 1210002. (in Chinese). doi: 10.3788/gzxb20174612.1210002 [14] DONG L M, YANG Q X, WU H Y, et al. High quality multi-spectral and panchromatic image fusion technologies based on curvelet transform[J]. Neurocomputing, 2015, 159: 268-274. doi: 10.1016/j.neucom.2015.01.050 [15] HUANG Y, BI D Y, WU D P. Infrared and visible image fusion based on different constraints in the non-subsampled shearlet transform domain[J]. Sensors, 2018, 18(4): 1169. doi: 10.3390/s18041169 [16] WANG Z Y, LI X F, DUAN H R, et al. Medical image fusion based on convolutional neural networks and non-subsampled contourlet transform[J]. Expert Systems with Applications, 2021, 171: 114574. doi: 10.1016/j.eswa.2021.114574 [17] CH M M I, GHAFOOR A, BAKHSHI A D, et al. Medical image fusion using non subsampled contourlet transform and iterative joint filter[J]. Multimedia Tools and Applications, 2022, 81(3): 4495-4509. doi: 10.1007/s11042-021-11753-8 [18] 李文, 叶坤涛, 李晟. 基于优化PCNN与区域特征引导法则的图像融合[J]. 激光与红外,2021,51(8):1104-1112.LI W, YE K T, LI SH. Image fusion based on optimized PCNN and region feature guided rule[J]. Laser & Infrared, 2021, 51(8): 1104-1112. (in Chinese). [19] 杨艳春, 李永萍, 党建武, 等. 基于快速交替引导滤波和CNN的红外与可见光图像融合[J]. 光学 精密工程,2023,31(10):1548-1562. doi: 10.37188/OPE.20233110.1548YANG Y CH, LI Y P, DANG J W, et al. Infrared and visible image fusion based on fast alternating guided filtering and CNN[J]. Optics and Precision Engineering, 2023, 31(10): 1548-1562. (in Chinese). doi: 10.37188/OPE.20233110.1548 [20] WANG K P, QI G Q, ZHU ZH Q, et al. A novel geometric dictionary construction approach for sparse representation based image fusion[J]. Entropy, 2017, 19(7): 306. doi: 10.3390/e19070306 [21] LV X B, LI Y W, ZHU SH SH, et al. Snapshot spectral polarimetric light field imaging using a single detector[J]. Optics Letters, 2020, 45(23): 6522-6525. doi: 10.1364/OL.409476 [22] LI X B, HU H F, GOUDAIL F, et al. Fundamental precision limits of full Stokes polarimeters based on DoFP polarization cameras for an arbitrary number of acquisitions[J]. Optics Express, 2019, 27(22): 31261-31272. doi: 10.1364/OE.27.031261 [23] 杨之文, 高胜钢, 王培纲. 几种地物反射光的偏振特性[J]. 光学学报,2005,25(2):241-245. doi: 10.3321/j.issn:0253-2239.2005.02.022YANG ZH W, GAO SH J, WANG P G. Polarization of reflected light by earth objects[J]. Acta Optica Sinica, 2005, 25(2): 241-245. (in Chinese). doi: 10.3321/j.issn:0253-2239.2005.02.022 [24] DA CUNHA A L, ZHOU J, DO M N. The nonsubsampled contourlet transform: theory, design, and applications[J]. IEEE Transactions on Image Processing, 2006, 15(10): 3089-3101. doi: 10.1109/TIP.2006.877507 [25] OLSHAUSEN B A, FIELD D J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J]. Nature, 1996, 381(6583): 607-609. doi: 10.1038/381607a0 [26] 丁贵鹏, 陶钢, 李英超, 等. 基于非下采样轮廓波变换与引导滤波器的红外及可见光图像融合[J]. 兵工学报,2021,42(9):1911-1922.DING G P, TAO G, LI Y CH, et al. Infrared and visible images fusion based on non-subsampled contourlet transform and guided filter[J]. Acta Armamentarii, 2021, 42(9): 1911-1922. (in Chinese). [27] 许廷发, 李俊涛, 张一舟, 等. 真彩色传递双波段图像融合[J]. 中国光学,2014,7(3):402-410.XU T F, LI J T, ZHANG Y ZH, et al. True color transfer for dual band image fusion[J]. Chinese Optics, 2014, 7(3): 402-410. (in Chinese). -
下载:
点击查看大图
计量
- 文章访问数: 61
- HTML全文浏览量: 25
- PDF下载量: 12
- 被引次数: 0
