Super-resolution reconstruction of rock CT images based on Real-ESRGAN
-
摘要: 图像采集设备和地质环境等因素导致岩石CT图像分辨率低、细节不清晰，而现有图像超分辨率重建方法在表征内部高密度矿物质颗粒和孔裂隙时容易丢失细节。针对上述问题，采用改进的增强型超分辨率生成对抗网络（Real−ESRGAN）对岩石CT图像进行超分辨率重建。选取山西晋城无烟煤矿业集团有限责任公司赵庄煤矿15号煤层底板的砂岩为研究对象，研究不同图像放大倍数下Real−ESRGAN的重建性能，并将其与超分辨率卷积神经网络（SRCNN）、超分辨率生成对抗网络（SRGAN）、增强型超分辨率生成对抗网络（ESRGAN）、增强的深度超分辨率网络（EDSR）等算法进行对比。试验结果表明：① 使用Real−ESRGAN重建的高分辨率图像在视觉效果上比原始CT图像更清晰，重建图像中裂隙轮廓和高密度矿物质颗粒更加突出，图像可视性得到了极大提高。② 在客观评估方面，Real−ESRGAN算法在2倍超分辨率重建后图像的峰值信噪比（PSNR）高达36.880 dB，结构相似性（SSIM）达0.933。但随着放大倍数的增加，6倍超分辨率重建图像上的孔隙出现模糊，PSNR降至32.781 dB，SSIM为0.896。③ Real−ESRGAN重建超分辨图像的孔隙率和喉道长度分布占比与原始CT图像相比非常接近，保留了岩石重要的细观结构信息。Abstract: Due to factors such as image acquisition equipment and geological environment, rock CT images have low resolution and unclear details. However, existing image super-resolution reconstruction methods are prone to losing details when characterizing high-density mineral particles and pores and cracks inside. To solve the above problems, an improved enhanced super-resolution generative adversarial network (Real-ESRGAN) is used for super-resolution reconstruction of rock CT images. The sandstone of the 15th coal seam floor in Zhaozhuang Coal Mine, Shanxi Jincheng Anthracite Mining Group Co., Ltd. is selected as the research object to study the reconstruction performance of Real-ESRGAN under different image magnifications. It is compared with algorithms such as super-resolution convolutional neural network (SRCNN), super-resolution generative adversarial network (SRGAN), enhanced super-resolution generative adversarial network (ESRGAN), and enhanced deep super-resolution network (EDSR). The experimental results show the following points. ① The high-resolution images reconstructed using Real-ESRGAN have clearer visual effects than the original CT images. The contours of cracks and high-density mineral particles in the reconstructed images are more prominent, greatly improving the visibility of the images. ② In terms of objective evaluation, the Real-ESRGAN algorithm achieves a peak signal-to-noise ratio (PSNR) of 36.880 dB and a structural similarity (SSIM) of 0.933 in the image after 2x super-resolution reconstruction. But as the magnification increases, the pores on the 6x super-resolution reconstructed image become blurry, with PSNR decreasing to 32.781 dB and SSIM reaching 0.896. ③ The porosity and throat length distribution ratio of the Real-ESRGAN reconstructed super-resolution image are very close to the original CT image, preserving important microstructural information of the rock.
-
图 3 原始图像与不同倍数低分辨率图像对比
Figure 3. Comparison between original image and low resolution images of different multiples
图 4 不同算法2倍超分辨率重建结果对比
Figure 4. Comparison of 2x super-resolution reconstruction results of different algorithms
图 5 不同算法4倍超分辨率重建结果对比
Figure 5. Comparison of 4x super-resolution reconstruction results of different algorithms
图 6 不同算法6倍超分辨率重建结果对比
Figure 6. Comparison of 6x super-resolution reconstruction results of different algorithms
图 7 不同算法2倍超分辨率单幅图像重建结果对比
Figure 7. Comparison of 2x super-resolution reconstruction results of a single image of different algorithms
图 8 Real−ESRGAN重建图像与原始CT图像孔隙率对比
Figure 8. Comparison of porosity between Real-ESRGAN reconstructed images and original CT images
图 9 不同算法重建图像与原始CT图像孔隙率对比
Figure 9. Comparison of porosity between reconstructed images of different algorithms and original CT images
图 10 Real−ESRGAN重建图像与原始CT图像喉道长度分布占比对比
Figure 10. Comparison of throat length distribution ratio between Real-ESRGAN reconstructed images and original CT images
图 11 不同算法重建图像与原始CT图像喉道长度分布占比对比
Figure 11. Comparison of throat length distribution ratio between reconstructed images of different algorithms and original CT images
表 1 不同算法超分辨率重建图像PSNR比较
Table 1. Comparison of PSNR of super-resolution reconstruction images with different algorithms
放大倍数 SRCNN SRGAN EDSR ESRGAN Real−ESRGAN 2 34.653 35.592 36.865 36.871 36.880 4 32.452 34.429 35.735 35.736 35.742 6 29.236 30.513 32.769 32.773 32.781 
下载: 导出CSV
表 2 不同算法超分辨率重建图像SSIM比较
Table 2. Comparison of SSIM of super-resolution reconstruction images with different algorithms
放大倍数 SRCNN SRGAN EDSR ESRGAN Real−ESRGAN 2 0.896 0.907 0.920 0.924 0.933 4 0.875 0.881 0.905 0.910 0.917 6 0.855 0.868 0.875 0.891 0.896
下载: 导出CSV
-
[1] 张艳博,徐跃东,刘祥鑫,等. 基于CT的岩石三维裂隙定量表征及扩展演化细观研究[J]. 岩土力学,2021,42(10):2659-2671.ZHANG Yanbo,XU Yuedong,LIU Xiangxin,et al. Quantitative characterization and mesoscopic study of propagation and evolution of three-dimensional rock fractures based on CT[J]. Rock and Soil Mechanics,2021,42(10):2659-2671. [2] ZHAO Yusong,GAO Yongtao,WU Shuchuan. Influence of different concealment conditions of parallel double flaws on mechanical properties and failure characteristics of brittle rock under uniaxial compression[J]. Theoretical and Applied Fracture Mechanics,2020,109(2). DOI: 10.1016/j.tafmec.2020.102751. [3] 杨琪,于岩斌,崔文亭,等. 单轴压缩下煤岩细观结构参数表征及演化规律[J]. 煤炭科学技术,2023,51(4):88-95.YANG Qi,YU Yanbin,CUI Wenting,et al. Fracture evolution of coal under uniaxial compression based on X-ray microscopic imaging[J]. Coal Science and Technology,2023,51(4):88-95. [4] 李文帅,王连国,陆银龙,等. 真三轴条件下砂岩强度、变形及破坏特征试验研究[J]. 采矿与安全工程学报,2019,36(1):191-197. doi: 10.13545/j.cnki.jmse.2019.01.025LI Wenshuai,WANG Lianguo,LU Yinlong,et al. Experimental investigation on the strength,deformation and failure characteristics of sandstone under true triaxial compression[J]. Journal of Mining & Safety Engineering,2019,36(1):191-197. doi: 10.13545/j.cnki.jmse.2019.01.025 [5] 张廷蓉,滕奇志,李征骥,等. 岩心三维CT图像超分辨率重建[J]. 浙江大学学报(工学版),2018,52(7):1294-1301.ZHANG Tingrong,TENG Qizhi,LI Zhengji,et al. Super-resolution reconstruction for three-dimensional core CT image[J]. Journal of Zhejiang University(Engineering Science),2018,52(7):1294-1301. [6] ZHU Shuyuan,ZENG Bing,ZENG Liaoyuan,et al. Image interpolation based on non-local geometric similarities and directional gradients[J]. IEEE Transactions on Multimedia,2016,18(9):1707-1719. doi: 10.1109/TMM.2016.2593039 [7] PAPYAN V,ELAD M. Multi-scale patch-based image restoration[J]. IEEE Transactions on Image Processing,2016,25(1):249-261. doi: 10.1109/TIP.2015.2499698 [8] DONG Chao,LOY C C,TANG Xiaoou. Accelerating the super-resolution convolutional neural network[C]. 14th European Conference on Computer Vision,Amsterdam,2016:391-407. [9] WANG Yingda,ARMSTRONG R T,MOSTAGHIMI P,et al. Enhancing resolution of digital rock images with super resolution convolutional neural networks[J]. Journal of Petroleum Science and Engineering,2019,182. DOI: 10.1016/j.petrol.2019.106261. [10] WANG Yukai,TENG Qizhi,HE Xiaohai,et al. CT-image of rock samples super resolution using 3D convolutional neural network[J]. Computers & Geosciences,2019,133. DOI: 10.1016/j.cageo.2019.104314. [11] DONG Chao,LOY C C,HE Kaiming,et al. Image super-resolution using deep convolutional networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(2):295-307. doi: 10.1109/TPAMI.2015.2439281 [12] LEDIG C,THEIS L,HUSZAR F,et al. Photo-realistic single image super-resolution using a generative adversarial network[C]. IEEE Conference on Computer Vision and Pattern Recognition,New York,2017:4681-4690. [13] GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al. Generative adversarial nets[J]. Communications of the ACM,2014,63(11):139-144. [14] WANG Xiantao,YU Ke,WU Shixiang,et al. Esrgan:enhanced super-resolution generative adversarial networks[C]. European Conference on Computer Vision Workshops,Munich,2019:63-79. [15] 辛元雪,朱凤婷,史朋飞,等. 基于改进增强型超分辨率生成对抗网络的图像超分辨率重建算法[J]. 激光与光电子学进展,2022,59(4):381-391.XIN Yuanxue,ZHU Fengting,SHI Pengfei,et al. Super-resolution reconstruction algorithm of images based on improved enhanced super-resolution generative adversarial network[J]. Laser & Optoelectronics Progress,2022,59(4):381-391. [16] LIM B,SON S,KIM H,et al. Enhanced deep residual networks for single image super-resolution[C]. IEEE Conference on Computer Vision and Pattern Recognition Workshops,Honolulu,2017:1132-1140. [17] WANG Xintao,XIE Liangbin,DONG Chao,et al. Real-ESRGAN:training real-world blind super-resolution with pure synthetic data[C]. IEEE/CVF International Conference on Computer Vision Workshops,Montreal,2021:1905-1914. [18] 方玉明,眭相杰,鄢杰斌,等. 无参考图像质量评价研究进展[J]. 中国图象图形学报,2021,26(2):265-286. doi: 10.11834/jig.200274FANG Yuming,SUI Xiangjie,YAN Jiebin,et al. Progress in no-reference image quality assessment[J]. Journal of Image and Graphics,2021,26(2):265-286. doi: 10.11834/jig.200274 [19] IOFFE,S,SZEGEDY,C. Batch normalization:accelerating deep network training by reducing internal covariate shift[C]. International Conference on Machine Learning,Lille,2015:448-456. [20] WANG Zhou,BOVIK A C,SHEIKH H R,et al. Image quality assessment:from error visibility to structural similarity[J]. IEEE Transactions on Image Processing,2004,13(4):600-612. doi: 10.1109/TIP.2003.819861 [21] 贾洁. 基于生成对抗网络的人脸超分辨率重建及识别[D]. 成都:电子科技大学,2018.JIA Jie. Face super-resolution reconstruction based on generative adversarial nets and face recognition[D]. Chengdu:University of Electronic Science and Technology,2018. [22] 朱新山,姚思如,孙彪,等. 图像质量评价:融合视觉特性与结构相似性指标[J]. 哈尔滨工业大学学报,2018,50(5):121-128. doi: 10.11918/j.issn.0367-6234.20180517ZHU Xinshan,YAO Siru,SUN Biao,et al. Image quality assessment:combining the characteristics of HVS and structural similarity index[J]. Journal of Harbin Institute of Technology,2018,50(5):121-128. doi: 10.11918/j.issn.0367-6234.20180517 [23] 雷健,潘保芝,张丽华. 基于数字岩心和孔隙网络模型的微观渗流模拟研究进展[J]. 地球物理学进展,2018,33(2):653-660. doi: 10.6038/pg2018BB0108LEI Jian,PAN Baozhi,ZHANG Lihua. Advance of microscopic flow simulation based on digital cores and pore network[J]. Progress in Geophysics,2018,33(2):653-660. doi: 10.6038/pg2018BB0108 -
点击查看大图
计量
- 文章访问数: 1397
- HTML全文浏览量: 65
- PDF下载量: 49
- 被引次数: 0
