基于结构纹理分解的矿井图像增强方法

张红; 索霆锋; 宋婉莹

doi:10.13272/j.issn.1671-251x.2023100005

基于结构纹理分解的矿井图像增强方法

doi: 10.13272/j.issn.1671-251x.2023100005

西安科技大学通信与信息工程学院，陕西西安　710054

基金项目: 国家自然科学基金资助项目（61901358）。

详细信息

作者简介:
张红（1972—），女，陕西西安人，副教授，研究方向为信息论、图像处理，E-mail：zhanghong@xust.edu.cn

通讯作者:
索霆锋（1998—），男，陕西宝鸡人，硕士研究生，研究方向为计算机视觉，E-mail：suooo1127@qq.com。

中图分类号: TD67
计量
- 文章访问数: 49
- HTML全文浏览量: 19
- PDF下载量: 7
- 被引次数: 0
出版历程
- 收稿日期: 2023-06-04
- 修回日期: 2024-03-06
- 网络出版日期: 2024-03-18

A mine image enhancement method based on structural texture decomposition

College of Communication and Information Engineering, Xi'an University of Science and Technology, Xi'an 710054, China

摘要

摘要: 矿井下存在低照度、多灰尘现象，导致监控视频采集的图像具有光照不均、模糊及细节丢失的问题，影响后续智能图像识别，现有矿井图像增强方法普遍存在图像纹理细节不清晰、视觉效果差的问题。提出了一种基于结构纹理分解的图像增强方法。首先，利用maxRGB算法对原始图像提取初始光照分量，接着构建优化目标函数，依次优化求解初始光照分量中的结构分量、纹理分量及噪声分量：先对初始光照分量进行加权引导滤波，作为先验约束，迭代获得边缘清晰的结构分量；再结合最大邻域差方法和加权平均局部变分构建局部变化偏差函数，作为约束权重，迭代得到细节丰富的纹理分量。然后，将原始图像转换到HSV颜色空间，提取出原始图像的亮度分量，并结合结构分量、纹理分量及噪声分量，利用Retinex理论进行重构，得到增强后的初始亮度分量。为避免亮度过增强，引入带有截断因子的自适应伽马校正（AGCWD）处理图像初始亮度信息，以获得最终的亮度分量。最后，将图像转换到RGB颜色空间，得到增强图像。实验结果表明：① 基于结构纹理分解的图像增强算法能保证图像边缘纹理细节更加清晰，减少了图像增强过程中的光晕伪影，且增强后的图像灰度直方图更均衡。② 与结构纹理感知Retinex（STAR）算法、联合内外先验（JieP）算法、加权变分模型（WVM）、半解耦分解（SDD）算法、带色彩恢复的多尺度Retinex（MSRCR）算法等5种图像增强算法相比，基于结构纹理分解的图像增强算法的自然图像质量评价指标（NIQE）分别降低了8.69%，29.05%，11.2%，29.53%，33.54%，视觉质量保真度（VIF）分别提高了91.17%，117.86%，59.38%，48.78%，183.12%，信息熵指标（Entropy）分别提高了3.20%，8.02%，4.07%，3.49%，22.68%。③ 基于结构纹理分解的图像增强算法运行时间仅长于MSRCR算法，但增强效果更好，能够满足矿井下图像增强的需求。
- 矿井图像增强 /
- 结构纹理分解 /
- 变分模型 /
- Retinex理论 /
- 自适应伽马校正
Abstract: There is a phenomenon of low lighting and excessive dust in underground mines, which leads to uneven lighting, blurriness, and loss of details in the images captured by monitoring videos. It affects subsequent intelligent image recognition. Existing mine image enhancement methods generally suffer from unclear texture details and poor visual effects. A method for image enhancement based on structural texture decomposition is proposed. Firstly, the maxRGB algorithm is used to extract the initial lighting component from the original image. Then, an optimization objective function is constructed to sequentially optimize and solve the structural component, texture component, and noise component in the initial lighting component. The weighted guided filtering is applied to the initial lighting component as a prior constraint, and the structural component with clear edges is obtained iteratively. Combined with the maximum neighborhood difference method and weighted average local variation, a local variation deviation function is constructed as constraint weights. The texture component with rich details is obtained iteratively. Secondly, the original image is transformed into the HSV color space, and the lighting component of the original image is extracted. Combined with the structural component, texture component, and noise component, Retinex theory is used for reconstruction to obtain the enhanced initial lighting component. To avoid excessive lighting enhancement, adaptive Gamma correction with weight distribution (AGCWD) is introduced to process the initial lighting information of the image and obtain the final lighting component. Finally, the image is converted to RGB color space to obtain an enhanced image. The experimental results show the following points. ① The image enhancement algorithm based on structural texture decomposition can ensure clearer texture details at the edges of the image, reduce halo artifacts during the image enhancement process, and achieve a more balanced grayscale histogram of the enhanced image. ② Compared with five image enhancement algorithms, including the Retinex algorithm based on structure and texture aware Retinex(STAR), the joint intrinsic-extrinsic prior model (JieP), the weighted variational mode (WVM), the semi-decoupled decomposition (SDD), and multi-scale Retinex with color restoration (MSRCR), the natural image quality evaluator (NIQE) of the image enhancement algorithm based on structural texture decomposition is reduced by 8.69%, 29.05%, 11.2%, 29.53%, and 33.54%, respectively. The visual information fidelity (VIF) increases by 91.17%, 117.86%, 59.38%, 48.78%, 183.12%, and the entropy index (Entropy) increases by 3.20%, 8.02%, 4.07%, 3.49%, and 22.68%, respectively. ③ The image enhancement algorithm based on structural texture decomposition has a running time only longer than the MSRCR algorithm. But the enhancement effect is better, which can meet the needs of image enhancement in underground mines.
- mine image enhancement /
- structural texture decomposition /
- variational model /
- Retinex theory /
- adaptive Gamma correction

HTML全文

图 1 基于结构纹理分解的矿井图像增强算法总体框架

Figure 1. Overall framework of mine image enhancement algorithm based on structural texture decomposition

下载: 全尺寸图片幻灯片

图 2 变分模型的求解步骤

Figure 2. Solution steps of variational model

下载: 全尺寸图片幻灯片

图 3 矿井图像的结构纹理分解

Figure 3. Structural texture decomposition of mine images

下载: 全尺寸图片幻灯片

图 4 不同场景下的矿井低照度数据集

Figure 4. Low lighting datasets for mines in different scenarios

下载: 全尺寸图片幻灯片

图 5 不同参数组合下的平均客观指标

Figure 5. Average objective indicators under different parameter combinations

下载: 全尺寸图片幻灯片

图 6 矿井低照度图像增强结果

Figure 6. Enhancement results of low lighting images in mines

下载: 全尺寸图片幻灯片

图 7 场景1（采矿轨道）图像增强结果

Figure 7. Enhancement results of image in scenario 1 (mining track)

下载: 全尺寸图片幻灯片

图 8 场景2（采矿巷道）图像增强结果

Figure 8. Enhancement results of image in scenario 2 (mining roadway)

下载: 全尺寸图片幻灯片

图 9 场景3（采矿工作面）图像增强结果

Figure 9. Enhancement results of image in scenario 3 (mining face)

下载: 全尺寸图片幻灯片

图 10 场景4（运煤输送带）图像增强结果

Figure 10. Enhancement results of image in scenario 4 (coal conveyor belt)

下载: 全尺寸图片幻灯片

图 11 场景1下不同方法增强后的直方图

Figure 11. Histograms enhanced by different methods in scenario 1

下载: 全尺寸图片幻灯片

表 1 不同算法的客观指标对比

Table 1. Comparison of objective indicators of different algorithms

算法	NIQE				Entropy				VIF
算法	场景1	场景2	场景3	场景4	场景1	场景2	场景3	场景4	场景1	场景2	场景3	场景4
原图	2.67	3.29	3.21	2.31	6.41	5.89	5.36	4.91
STAR	2.47	2.34	2.41	1.97	7.29	7.16	7.07	7.19	1.17	3.88	3.88	5.11
JieP	2.57	2.38	2.60	4.31	7.25	7.10	6.90	6.20	1.59	3.51	3.34	3.87
WVM	2.81	2.50	2.21	1.90	7.13	7.09	7.31	6.93	2.04	5.22	4.50	5.06
SDD	2.53	3.20	2.77	3.43	7.33	7.25	7.06	7.01	2.56	5.03	5.25	5.19
MSRCR	2.73	2.84	3.55	3.51	6.41	5.89	5.36	6.51	2.41	2.35	1.38	3.33
本文	2.35	2.20	1.99	1.87	7.59	7.37	7.48	7.21	3.08	10.08	8.36	5.31

下载: 导出CSV

表 2 平均运行时间

Table 2. Average running times s

算法	STAR	JieP	WVM	SDD	MSRCR	本文
时间	16.21	16.40	38.16	19 .48	1.38	13.51

下载: 导出CSV

参考文献(21)

[1]	丁恩杰,俞啸,夏冰,等. 矿山信息化发展及以数字孪生为核心的智慧矿山关键技术[J]. 煤炭学报,2022,47(1):564-578. DING Enjie,YU Xiao,XIA Bing,et al. Development of mine informatization and key technologies of intelligent mines[J]. Journal of China Coal Society,2022,47(1):564-578.
[2]	王海军,王洪磊. 带式输送机智能化关键技术现状与展望[J]. 煤炭科学技术,2022,50(12):225-239. WANG Haijun,WANG Honglei. Status and prospect of intelligent key technologies of belt conveyor[J]. Coal Science and Technology,2022,50(12):225-239.
[3]	JOBSON D J,RAHMAN Z,MEMBER. Properties and performance of a center/surround retinex[J]. IEEE Transactions on Image Processing,1997,6(3):451-462. doi: 10.1109/83.557356
[4]	RAHMAN Z,JOBSON D J,WOODELL G A. Multi-scale retinex for color image enhancement[C]. 3rd IEEE International Conference on Image Processing,Lausanne,1996 :1003-1006.
[5]	JOBSON D J,RAHMAN Z,WOODELL G A. A multiscale retinex for bridging the gap between color images and the human observation of scenes[J]. IEEE Transactions on Image Processing,1997,7(6):965-976.
[6]	张立亚,郝博南,孟庆勇,等. 基于HSV空间改进融合Retinex算法的井下图像增强方法[J]. 煤炭学报,2020,45(增刊1):532-540. ZHANG Liya,HAO Bonan,MENG Qingyong,et al. Method of image enhancement in coal mine based on improved retex fusion algorithm in HSV space[J]. Journal of China Coal Society,2020,45(S1):532-540.
[7]	唐守锋,史可,仝光明,等. 一种矿井低照度图像增强算法[J]. 工矿自动化,2021,47(10):32-36. TANG Shoufeng,SHI Ke,TONG Guangming,et. al. A mine low illumination image enhancement algorithm[J]. Industry and Mine Automation,2021,47(10):32-36.
[8]	洪炎,朱丹萍,龚平顺. 基于TopHat加权引导滤波的Retinex矿井图像增强算法[J]. 工矿自动化,2022,48(8):43-49. HONG Yan,ZHU Danping,GONG Pingshun. Retinex mine image enhancement algorithm based on TopHat weighted guided filtering[J]. Journal of Mine Automation,2022,48(8):43-49.
[9]	龚云,颉昕宇. 基于同态滤波方法的煤矿井下图像增强技术研究[J]. 煤炭科学技术,2023,51(3):241-250. GONG Yun,JIE Xinyu. Research on coal mine underground image recognition technology based on homomorphic filtering method[J]. Coal Science and Technology,2023,51(3):241-250.
[10]	龙庆延,王正勇,潘建,等. 基于奇异值分解和引导滤波的低照度图像增强算法[J]. 科学技术与工程,2021,21(12):5018-5023. LONG Qingyan,WANG Zhengyong,PAN Jian,et al. Low light image enhancement algorithm based on singular value decomposition and guided filtering[J]. Science Technology and Engineering,2021,21(12):5018-5023.
[11]	LI Zhengguo,ZHENG Jinghong,ZHU Zijian,et al. Weighted guided image filtering[J]. IEEE Transactions on Image Processing,2015,24(1):120-129. doi: 10.1109/TIP.2014.2371234
[12]	ZHANG Xiaoting,HE Chuanjiang. Robust double-weighted guided image filtering[J]. Signal Processing,2022,199. DOI: 10.1016/J.SIGPRO.2022.108609.
[13]	CHIEN C,KINOSHITA Y,KIYA H. A noise-aware enhancement method for underexposed images[C]. IEEE International Conference on Consumer Electronics Asia ,Bangkok,2019:131-134.
[14]	KHAN R,MEHMOOD A,ZHENG Zhonglong. Robust contrast enhancement method using a retinex model with adaptive brightness for detection applications[J]. Optics Express,2022,30(21):37736-37752. doi: 10.1364/OE.472557
[15]	FU Xueyang,ZENG Delu,HUANG Yue,et al. A weighted variational model for simultaneous reflectance and illumination estimation[C]. IEEE Conference on Computer Vision and Pattern Recognition,Las Vegas,2016:2782-2790.
[16]	SHEIKH H R,BOVIK A C. Image information and visual quality[J]. IEEE Transactions on Image Processing,2006,15(2):430-444. doi: 10.1109/TIP.2005.859378
[17]	MITTAL A,SOUNDARARAJAN R,BOVIK A C. Making a "completely blind" image quality analyzer[J]. IEEE Signal Processing Letters,2013,20(3):209-212. doi: 10.1109/LSP.2012.2227726
[18]	GUO Xiaojie,LI Yu,LING Haibin. LIME:low-light image enhancement via illumination map estimation[J]. IEEE Transactions on Image Processing,2017,26(2):982-993. doi: 10.1109/TIP.2016.2639450
[19]	XU Jun,HOU Yingkun,REN Dongwei,et al. STAR:a structure and texture aware Retinex model[J]. IEEE Transactions on Image Processing,2020,29:5022-5037. doi: 10.1109/TIP.2020.2974060
[20]	CAI Bolun,XU Xianming,GUO Kailing,et al. A joint intrinsic-extrinsic prior model for Retinex[C]. IEEE International Conference on Computer Vision,Venice,2017:4000-4009.
[21]	HAO Shijie,HAN Xu,GUO Yanrong,et al. Lowlight image enhancement with semi-decoupled decomposition[J]. IEEE Transactions on Multimedia,2020,22(12):3025-3038. doi: 10.1109/TMM.2020.2969790