基于改进引导滤波器的多光谱去马赛克方法

齐海超; 宋延嵩; 张博; 梁宗林; 闫纲琦; 薛佳音; 张轶群; 任斌

doi:10.37188/CO.2022-0231

基于改进引导滤波器的多光谱去马赛克方法

doi: 10.37188/CO.2022-0231

齐海超^1,,
宋延嵩^{1, 2, ,},
张博¹,
梁宗林¹,
闫纲琦¹,
薛佳音²,
张轶群²,
任斌³

1.
长春理工大学光电工程学院空间光电技术研究所, 长春 130022
2.
鹏城实验室, 深圳 518052
3.
中国空间技术研究院西安分院, 西安 710212

基金项目: 国家重点研发计划资助项目（No. 2022YFB3902500）；国家自然科学基金资助项目（No. U2141231）；吉林省自然科学基金（No. 202002036JC）；鹏城实验室重大攻关项目（No. PCL2021A03-1）

详细信息

作者简介:
齐海超（1997—），男，吉林长春人，硕士研究生，2019年于长春理工大学获得学士学位，主要从事图像处理方面的研究。E-mail：qihaichao2019@163.com

宋延嵩(1983—)，男，吉林长春人，博士，研究员，博士生导师，2006年、2009年、2014年于长春理工大学分别获得学士、硕士及博士学位，主要研究方向为空间激光通信技术。E-mail: songyansong2006@126.com

中图分类号: TP391.9
计量
- 文章访问数: 483
- HTML全文浏览量: 318
- PDF下载量: 224
- 被引次数: 0
出版历程
- 收稿日期: 2022-11-13
- 修回日期: 2022-12-12
- 网络出版日期: 2023-04-17

Multispectral demosaicing method based on an improved guided filter

1.
Institute of Space Photoelectric Technology, School of Opto-Electronic Engineering, Changchun University of Science and Technology, Changchun 130022, China
2.
Peng Cheng Laboratory, Shenzhen 518052, China
3.
Xi’an Branch of China Academy of Space Technology, Xi’an 710212, China

Funds: Supported by National Key R & D Program of China (No. 2022YFB3902500); National Natural Science Foundation of China (No.U2141231); the Natural Science Foundation of Jilin Province (No. 202002036JC); The Major Key Project of PCL (No. PCL2021A03-1)

More Information

Corresponding author: songyansong2006@126.com

摘要

摘要:
为了更好地保留多光谱去马赛克图像中的高频信息，本文提出了一种基于改进引导滤波器的多光谱图像去马赛克方法。首先，基于自回归模型对相邻像素点间的强相关性进行建模，在每个像素处渐进估计其模型参数，通过最小化局部窗口内的估计误差，得到最优估计值来插值采样密集波段G，并生成高质量的引导图像；然后，引入加窗固有变分系数到惩罚因子中，得到具有边缘感知能力的加权引导滤波器并重建其余稀疏采样波段。最后，使用CAVE数据集和TokyoTech数据集进行仿真。实验结果表明：相较于主流的5波段多光谱图像去马赛克方法，本方法重建图像的峰值信噪比和结构相似度在CAVE数据集和TokyoTech数据集上分别提高了3.40%，2.02%，1.34%，0.30%和6.11%，5.95%，2.28%，1.42%，且更好地保留了原始图像的局部结构和颜色信息，减少了边缘伪影和噪声现象的出现。
- 计算成像 /
- 多光谱滤光片阵列 /
- 多光谱去马赛克方法 /
- 自回归模型 /
- 加权引导滤波器
Abstract:
In order to better preserve high-frequency information in demosaicing multispectral images, we propose a new demosaicing method for multispectral images based on an improved guided filter. Firstly, the strong correlation between adjacent pixels based on the autoregressive model is constructed, gradually estimates the model parameters at each pixel, and the optimal estimation value is obtained by minimizing the estimation error in the local window, interpolates the sampling dense band G, and generates high-quality guide images. The windowed intrinsic variation coefficient is then introduced into the penalty factor to obtain a weighted guide filter with edge sensing ability and to reconstruct the remaining sparse sampling bands. Finally, the CAVE dataset and the TokyoTech dataset are used for simulation. The experimental results show that compared with the mainstream five-band multispectral image demosaicing method, the peak signal-to-noise ratio and structure similarity of the reconstructed image in the CAVE dataset and the TokyoTech dataset are improved by 3.40%, 2.02%, 1.34%, 0.30% and 6.11%, 5.95%, 2.28%, 1.42%, respectively. The local structure and color information of the original image are also better preserved, and the edge artifacts and noise are reduced.
- computational imaging /
- multispectral filter array /
- multispectral demosaicing method /
- autoregressive model /
- weighted guided filter

HTML全文

图 1 5波段MSFA(a)二叉树分裂过程及(b)排列模式

Figure 1. (a) Binary tree splitting process and (b) arrangement of Five-band MSFA

下载: 全尺寸图片幻灯片

图 2 邻域T内像素排列

Figure 2. Pixel arrangement in neighborhood T

下载: 全尺寸图片幻灯片

图 3 水平-垂直方向上的自回归模型

Figure 3. Autoregressive model in the horizontal-vertical direction

下载: 全尺寸图片幻灯片

图 4 模型参数估计

Figure 4. Estimation of model parameter

下载: 全尺寸图片幻灯片

图 5 基于加权引导滤波的去马赛克流程

Figure 5. Demosaicing process based on weight-guided filtering

下载: 全尺寸图片幻灯片

图 6 不同算法Balloons场景重建图像对比

Figure 6. Comparison of Balloons images reconstructed by different algorithms

下载: 全尺寸图片幻灯片

图 7 不同算法CD场景重建图像对比

Figure 7. Comparison of CD images reconstructed by different algorithms

下载: 全尺寸图片幻灯片

图 8 不同算法Party场景重建图像对比

Figure 8. Comparison of party images reconstructed by different algorithms

下载: 全尺寸图片幻灯片

表 1 CAVE数据集上3种方法的客观评价指标

Table 1. Objective evaluation metrics of the three methods on the CAVE dataset

CAVE	sRGB PSNR/dB			sRGB SSIM			CIEDE 2000
CAVE	GF	APMID	Pro	GF	APMID	Pro	GF	APMID	Pro
Balloons	41.62	42.68	43.11	0.9859	0.9916	0.9936	1.18	1.06	0.99
Clay	37.33	37.63	38.69	0.8758	0.8817	0.8852	1.07	0.94	0.89
Beers	41.57	42.18	43.52	0.9816	0.9868	0.9894	1.25	1.28	1.07
Lemons	42.91	42.87	42.91	0.9749	0.9805	0.9822	1.11	1.08	1.03
Peppers	42.14	42.08	42.52	0.9715	0.9801	0.9805	0.89	0.78	0.73
Feathers	35.64	35.94	35.99	0.9413	0.9593	0.9614	2.30	2.10	2.04
Flowers	38.93	41.18	42.50	0.9523	0.9778	0.9824	1.26	0.91	0.83
Paints	36.16	34.88	36.32	0.9696	0.9729	0.9797	2.40	2.43	2.17
Apples	45.24	45.10	45.68	0.9838	0.9875	0.9886	0.77	0.75	0.70
Toys	38.83	41.29	42.77	0.9666	0.9845	0.9891	1.24	0.90	0.80
Avg	40.04	40.58	41.40	0.9603	0.9703	0.9732	1.35	1.22	1.13

下载: 导出CSV

表 2 TokyoTech数据集上3种方法的客观评价指标

Table 2. Objective evaluation metrics of the three methods on the TokyoTech dataset

TokyoTech	sRGB PSNR/dB			sRGB SSIM			CIEDE 2000
TokyoTech	GF	APMID	Pro	GF	APMID	Pro	GF	APMID	Pro
Butterfly	37.53	38.95	40.44	0.9596	0.9678	0.9810	1.60	1.42	1.17
Butterfly3	38.42	42.94	41.99	0.9487	0.9777	0.9793	1.37	0.91	0.84
Butterfly4	40.63	40.73	42.37	0.9691	0.9590	0.9827	1.12	1.23	0.87
CD	32.20	32.78	32.87	0.9450	0.9580	0.9629	1.97	1.72	1.65
Character	37.74	37.45	38.14	0.9673	0.9736	0.9835	1.83	1.94	1.73
Cloth	34.18	35.00	35.87	0.9321	0.9481	0.9573	3.34	3.22	2.75
Color	39.22	38.62	41.36	0.9782	0.9630	0.9895	1.74	2.00	1.47
Colorchart	42.83	44.79	47.80	0.9819	0.9847	0.9941	0.92	0.77	0.55
Fan2	32.68	33.29	34.09	0.9257	0.9426	0.9629	2.63	2.34	2.04
Party	32.79	33.45	35.78	0.9366	0.9509	0.9693	2.06	1.66	1.36
Avg	36.82	37.80	39.07	0.9544	0.9625	0.9762	1.86	1.72	1.44

下载: 导出CSV

表 3 不同方法在两种数据集上的运行时间

Table 3. Running times of different methods on the two datasets (s)

数据集	GF	APMID	Pro
CAVE	1.49	0.71	1.33
TokyoTech	1.36	0.56	1.29

下载: 导出CSV

参考文献(30)

[1]	ORTEGA S, HALICEK M, FABELO H, et al. Hyperspectral and multispectral imaging in digital and computational pathology: a systematic review[J]. Biomedical Optics Express, 2020, 11(6): 3195-3233. doi: 10.1364/BOE.386338
[2]	王成, 刘宾, 周楚, 等. 窄带LED照明的多光谱显微成像系统研究[J]. 中国激光,2020,47(12):1207006. doi: 10.3788/CJL202047.1207006 WANG CH, LIU B, ZHOU CH, et al. Multispectral microimaging system with narrowband LED illumination[J]. Chinese Journal of Lasers, 2020, 47(12): 1207006. (in Chinese) doi: 10.3788/CJL202047.1207006
[3]	唐凌宇, 葛明锋, 董文飞. 全自动推扫式高光谱显微成像系统设计与研究[J]. 中国光学,2021,14(6):1486-1494. doi: 10.37188/CO.2021-0040 TANG L Y, GE M F, DONG W F. Design and research of fully automatic push-broom hyperspectral microscopic imaging system[J]. Chinese Optics, 2021, 14(6): 1486-1494. (in Chinese) doi: 10.37188/CO.2021-0040
[4]	SU W H, SUN D W. Multispectral imaging for plant food quality analysis and visualization[J]. Comprehensive Reviews in Food Science and Food Safety, 2018, 17(1): 220-239. doi: 10.1111/1541-4337.12317
[5]	CHAMBINO L L, SILVA J S, BERNARDINO A. Multispectral facial recognition: a review[J]. IEEE Access, 2020, 8: 207871-207883. doi: 10.1109/ACCESS.2020.3037451
[6]	WU F, JING X Y, FENG Y J, et al. Spectrum-aware discriminative deep feature learning for multi-spectral face recognition[J]. Pattern Recognition, 2021, 111: 107632. doi: 10.1016/j.patcog.2020.107632
[7]	李云辉. 压缩光谱成像系统中物理实现架构研究综述[J]. 中国光学(中英文),2022,15(5):929-945. doi: 10.37188/CO.2022-0104 LI Y H. Review of physical implementation architecture in compressive spectral imaging system[J]. Chinese Optics, 2022, 15(5): 929-945. (in Chinese) doi: 10.37188/CO.2022-0104
[8]	杨鹰, 孔玲君, 刘真. 基于压缩感知的多光谱图像去马赛克算法[J]. 液晶与显示,2017,32(1):56-61. doi: 10.3788/YJYXS20173201.0056 YANG Y, KONG L J, LIU ZH. Multi-spectral demosaicking method based on compressive sensing[J]. Chinese Journal of Liquid Crystals and Displays, 2017, 32(1): 56-61. (in Chinese) doi: 10.3788/YJYXS20173201.0056
[9]	HABTEGEBRIAL T A, REIS G, STRICKER D. Deep convolutional networks for snapshot hypercpectral demosaicking[C]. 2019 10th Workshop on Hyperspectral Imaging and Signal Processing: Evolution in Remote Sensing (WHISPERS), IEEE, 2019: 1-5.
[10]	FENG K, ZHAO Y Q, CHAN J C W, et al. Mosaic convolution-attention network for demosaicing multispectral filter array images[J]. IEEE Transactions on Computational Imaging, 2021, 7: 864-878. doi: 10.1109/TCI.2021.3102052
[11]	肖树林, 胡长虹, 高路尧, 等. 像元映射变分辨率光谱成像重构[J]. 中国光学(中英文),2022,15(5):1045-1054. XIAO SH L, HU CH H, GAO L Y, et al. Pixel mapping variable-resolution spectral imaging reconstruction[J]. Chinese Optics, 2022, 15(5): 1045-1054. (in Chinese)
[12]	MIAO L D, QI H R. The design and evaluation of a generic method for generating mosaicked multispectral filter arrays[J]. IEEE Transactions on Image Processing, 2006, 15(9): 2780-2791. doi: 10.1109/TIP.2006.877315
[13]	MIAO L D, QI H R, RAMANATH R, et al. Binary tree-based generic demosaicking algorithm for multispectral filter arrays[J]. IEEE Transactions on Image Processing, 2006, 15(11): 3550-3558. doi: 10.1109/TIP.2006.877476
[14]	GUPTA M, RAM M. Weighted bilinear interpolation based generic multispectral image demosaicking method[J]. Journal of Graphic Era University, 2019, 7(2): 108-118.
[15]	GUPTA M, RATHI V, GOYAL P. Adaptive and progressive multispectral image demosaicking[J]. IEEE Transactions on Computational Imaging, 2022, 8: 69-80. doi: 10.1109/TCI.2022.3140554
[16]	孙帮勇, 袁年曾, 胡炳樑. 一种八谱段滤光片成像系统设计[J]. 光子学报,2020,49(5):0511001. doi: 10.3788/gzxb20204905.0511001 SUN B Y, YUAN N Z, HU B L. Design of an eight-band filter imaging system[J]. Acta Photonica Sinica, 2020, 49(5): 0511001. (in Chinese) doi: 10.3788/gzxb20204905.0511001
[17]	RATHI V, GOYAL P. Generic multispectral Demosaicking based on directional interpolation[J]. IEEE Access, 2022, 10: 64715-64728. doi: 10.1109/ACCESS.2022.3182493
[18]	MONNO Y, TANAKA M, OKUTOMI M. Multispectral demosaicking using guided filter[J]. Proceedings of SPIE, 2012, 8299: 82990O. doi: 10.1117/12.906168
[19]	任杰, 刘家瑛, 白蔚, 等. 基于隐式分段自回归模型的图像插值算法[J]. 软件学报,2012,23(5):1248-1259. doi: 10.3724/SP.J.1001.2012.04049 REN J, LIU J Y, BAI W, et al. Image interpolation algorithm based on implicit piecewise autoregressive model[J]. Journal of Software, 2012, 23(5): 1248-1259. (in Chinese) doi: 10.3724/SP.J.1001.2012.04049
[20]	ZHANG X J, WU X L. Image interpolation by adaptive 2-D autoregressive modeling and soft-decision estimation[J]. IEEE Transactions on Image Processing, 2008, 17(6): 887-896. doi: 10.1109/TIP.2008.924279
[21]	HE K M, SUN J, TANG X O. Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(6): 1397-1409. doi: 10.1109/TPAMI.2012.213
[22]	LI ZH G, ZHENG J H, ZHU Z J, et al. Weighted guided image filtering[J]. IEEE Transactions on Image Processing, 2015, 24(1): 120-129. doi: 10.1109/TIP.2014.2371234
[23]	XU L, YAN Q, XIA Y, et al. Structure extraction from texture via relative total variation[J]. ACM Transactions on Graphics, 2012, 31(6): 139.
[24]	路陆, 姜鑫, 杨锦程, 等. 基于自适应引导滤波的红外图像细节增强[J]. 液晶与显示,2022,37(9):1182-1189. doi: 10.37188/CJLCD.2022-0024 LU L, JIANG X, YANG J CH, et al. Adaptive guided filtering based infrared image detail enhancement[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(9): 1182-1189. (in Chinese) doi: 10.37188/CJLCD.2022-0024
[25]	YASUMA F, MITSUNAGA T, ISO D, et al. Generalized assorted pixel camera: Postcapture control of resolution, dynamic range, and spectrum[J]. IEEE Transactions on Image Processing, 2010, 19(9): 2241-2253. doi: 10.1109/TIP.2010.2046811
[26]	MONNO Y, KIKUCHI S, TANAKA M, et al. A practical one-shot multispectral imaging system using a single image sensor[J]. IEEE Transactions on Image Processing, 2015, 24(10): 3048-3059. doi: 10.1109/TIP.2015.2436342
[27]	PARK J I, LEE M H, GROSSBERG M D, et al. . Multispectral imaging using multiplexed illumination[C]. 2007 IEEE 11th International Conference on Computer Vision, IEEE, 2007: 1-8.
[28]	SARA U, AKTER M, UDDIN M S. Image quality assessment through FSIM, SSIM, MSE and PSNR—a comparative study[J]. Journal of Computer and Communications, 2019, 7(3): 8-18. doi: 10.4236/jcc.2019.73002
[29]	SHARMA G, WU W CH, DALAL E N. The CIEDE2000 color-difference formula: Implementation notes, supplementary test data, and mathematical observations[J]. Color Research &Application, 2005, 30(1): 21-30.
[30]	贾停停, 王慧琴, 王可, 等. 相位相关性增强的自适应低重叠率多光谱图像快速拼接算法[J]. 液晶与显示,2022,37(4):483-493. doi: 10.37188/CJLCD.2021-0294 JIA T T, WANG H Q, WANG K, et al. Adaptive low overlap multispectral image fast mosaic algorithm based on phase correlation enhancement[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(4): 483-493. (in Chinese) doi: 10.37188/CJLCD.2021-0294