基于边缘引导的纺织品纹样数字化修复方法

doi:10.13475/j.fzxb.20231005101

摘要/Abstract

摘要：

针对破损纺织品文物人工修复周期长、易造成二次破坏等问题,提出一种用改进的Criminisi算法修复破损传统纺织品纹样的数字化方法。该算法通过对缺失边缘进行重构还原纺织品纹样的结构,用于引导进一步的纹理修复。首先使用线性或二阶贝塞尔曲线拟合缺失边缘,以恢复结构;然后利用多分辨率图像中的结构信息计算更有效的优先权,确定当前待修复块;再在多分辨率图像中根据颜色、梯度和边界特征计算多个候选匹配块,从中选择最佳匹配块以减少选择过程中的随机性;最后通过分割复制的最佳匹配块进行填补,以减少对已知信息区域的覆盖,迭代完成全部破损区域的修复。实验结果表明,本文算法对存在较多结构破坏的真实纺织品彩色图像可快速实现较为自然的修复效果。

关键词: 纺织品, 文物修复, 图像修复, 多分辨率图像, 结构信息

Abstract:

Objective This study aims to further improve the Criminisi algorithm to effectively restore traditional Chinese textile patterns in the presence of damage. Given the complex nature of these patterns, structural restoration is essential to ensure their accurate recovery. Efforts will be directed towards improving the inpainting algorithm's performance and expanding its applicability in practical textile restoration work, with the aim of achieving faster and more accurate repair results. Our research aims to provide more feasible solutions for the preservation of cultural heritage and textile restoration, ensuring the enduring legacy of traditional Chinese textile patterns.

Method Firstly, it uses linear or second-order BÉzier curves to fit the missing edges and restore the structure. Then, it calculates more effective priority using structural information from a multi-resolution image to determine the current patch to be repaired. Next, it computes multiple candidates matching patches in the multi-resolution image based on color, gradient, and boundary features, selecting the best-matched patch to reduce randomness in the selection process. Finally, the replicated best-matched patch is segmented before being used for filling the damaged area, reducing the overlap with known information regions and achieving iterative completion of the restoration for all damaged areas.

Results Real traditional textile images were collected, and artificial damage was introduced by adding masks to simulate challenges encountered when dealing with damaged textiles. The effectiveness of the proposed algorithm was evaluated using a variety of objective metrics, including the peak signal-to-noise ratio, structural similarity feature similarity index measure, feature similarity index measure, and edge preservation rate. These metrics provided a comprehensive and quantifiable assessment of the restoration results. Apart from the quantitative assessments, a subjective evaluation of the inpainting results was also carried out. This qualitative assessment revealed that the proposed algorithm excelled in fitting the main structures in the damaged areas, ultimately resulting in a more visually pleasing restoration effect. Experimental results demonstrate that our method achieves higher objective evaluation scores and natural restoration effects for real textile color images with significant structural damages. It is worth noting that, despite the superior inpainting quality achieved by the proposed algorithm, there was a trade-off in terms of the time required for restoration. Nevertheless, this minor sacrifice in time was considered acceptable, given the significant enhancement in inpainting quality.

Conclusion Digital restoration of textile artifacts is crucial for preserving our cultural heritage. It protects these artifacts from decay and loss, allowing researchers and educators to explore their historical and artistic value. Digital restoration enables easy sharing and accessibility of artifacts, both online and physically, reaching a broader audience. Its speed compared to manual restoration expedites the process, swiftly presenting artifacts in their original aesthetic state for appreciation and study. Moreover, it helps preserve historical information within artifacts, contributing to the revitalization of their aesthetic value and fostering a deeper recognition of their artistic significance.

Key words: textile, cultural relics restoration, image inpainting, multi-resolution image, structural information

中图分类号:

TS941.26

张婧, 辛斌杰, 袁智杰, 许颖琦. 基于边缘引导的纺织品纹样数字化修复方法[J]. 纺织学报, 2024, 45(02): 101-111.

ZHANG Jing, XIN Binjie, YUAN Zhijie, XU Yingqi. Digitized restoration of textile pattern through edge-guided image inpainting method[J]. Journal of Textile Research, 2024, 45(02): 101-111.

图/表 14

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

表1

不同算法的客观评价分数"

样本编号	方法	PSNR	SSIM	FSIM	EPRa	样本编号	方法	PSNR	SSIM	FSIM	EPRa
1	文献[7]	30.672 3	0.981 9	0.978 7	0.966 0	6	文献[7]	29.805 3	0.968 0	0.971 9	0.970 3
	文献[12]	27.318 2	0.976 7	0.969 7	0.959 9		文献[12]	29.149 5	0.966 2	0.963 3	0.954 6
	文献[23]	31.609 3	0.983 2	0.980 4	0.964 7		文献[23]	29.325 7	0.967 7	0.970 4	0.970 5
	文献[22]	31.881 0	0.985 0	0.980 0	0.963 4		文献[22]	31.329 5	0.971 8	0.975 7	0.964 1
	文献[15]	29.319 5	0.975 3	0.967 6	0.939 5		文献[15]	27.814 1	0.943 3	0.951 8	0.949 9
	本文算法	34.072 6	0.988 5	0.985 3	0.969 2		本文算法	31.730 7	0.973 5	0.977 9	0.973 1
2	文献[7]	30.327 4	0.983 7	0.980 4	0.979 1	7	文献[7]	27.174 6	0.981 8	0.960 1	0.949 5
	文献[12]	29.379 2	0.981 3	0.975 4	0.976 9		文献[12]	28.561 0	0.984 3	0.967 5	0.975 8
	文献[23]	30.076 9	0.983 2	0.980 6	0.979 4		文献[23]	29.534 3	0.983 7	0.971 0	0.977 3
	文献[22]	31.435 0	0.984 9	0.983 7	0.980 0		文献[22]	26.500 0	0.976 8	0.956 1	0.977 2
	文献[15]	28.279 5	0.972 9	0.965 4	0.964 5		文献[15]	28.893 8	0.984 1	0.968 9	0.968 9
	本文算法	32.673 5	0.986 7	0.984 8	0.979 6		本文算法	29.621 8	0.984 1	0.973 3	0.982 7
3	文献[7]	24.351 2	0.936 7	0.938 4	0.891 8	8	文献[7]	28.142 3	0.984 2	0.981 0	0.980 1
	文献[12]	20.045 5	0.914 6	0.917 5	0.457 1		文献[12]	28.008 9	0.982 6	0.981 1	0.980 2
	文献[23]	23.969 2	0.936 6	0.936 8	0.828 3		文献[23]	28.245 0	0.984 5	0.982 1	0.979 8
	文献[22]	23.712 3	0.932 4	0.933 2	0.804 2		文献[22]	28.678 0	0.985 7	0.981 8	0.980 1
	文献[15]	25.127 5	0.956 3	0.939 1	0.881 7		文献[15]	27.148 0	0.978 0	0.974 8	0.973 3
	本文算法	26.037 2	0.953 5	0.947 4	0.900 2		本文算法	30.716 0	0.989 6	0.985 2	0.982 0
4	文献[7]	24.233 3	0.956 4	0.937 3	0.873 2	9	文献[7]	30.993 0	0.985 3	0.983 2	0.976 6
	文献[12]	24.070 1	0.955 0	0.937 5	0.952 9		文献[12]	30.856 5	0.984 0	0.983 7	0.975 8
	文献[23]	26.222 9	0.959 0	0.952 2	0.956 8		文献[23]	31.213 1	0.985 1	0.984 2	0.976 8
	文献[22]	26.333 5	0.958 3	0.954 7	0.955 9		文献[22]	32.377 2	0.986 8	0.985 2	0.977 0
	文献[15]	25.167 3	0.941 4	0.941 6	0.939 9		文献[15]	29.551 4	0.972 1	0.975 6	0.952 9
	本文算法	27.816 4	0.964 0	0.964 7	0.959 1		本文算法	33.375 9	0.987 5	0.988 8	0.979 0
5	文献[7]	31.037 5	0.981 8	0.986 5	0.980 6	10	文献[7]	30.170 8	0.988 3	0.983 6	0.984 5
	文献[12]	28.979 7	0.977 0	0.977 1	0.976 7		文献[12]	28.643 4	0.986 3	0.982 4	0.981 3
	文献[23]	30.505 7	0.979 4	0.985 5	0.978 5		文献[23]	32.161 0	0.989 9	0.987 3	0.981 5
	文献[22]	30.849 5	0.980 7	0.985 2	0.980 2		文献[22]	32.957 3	0.990 8	0.991 1	0.984 7
	文献[15]	29.146 6	0.966 7	0.978 4	0.960 8		文献[15]	33.299 3	0.984 0	0.991 2	0.886 4
	本文算法	31.294 9	0.983 1	0.986 8	0.981 6		本文算法	37.783 9	0.993 1	0.997 0	0.985 6

表1

图12

图13

参考文献 28

[1]	BERTALMIO M, SAPIRO G, CASELLES V, et al. Image inpainting[C]// Proceedings of the 27th Annual Conference On Computer Graphics And Interactive Techniques. New York: ACM, 2000: 417-424.
[2]	SHEN J, CHAN T F. Mathematical models for local nontexture inpaintings[J]. SIAM Journal on Applied Mathematics, 2001, 62(3): 1019-1043. doi: 10.1137/S0036139900368844
[3]	RUDIN L I, OSHER S, FATEMI E. Nonlinear total variation based noise removal algorithms[J]. Physica D: Nonlinear Phenomena, 1992, 60(1-4): 259-268. doi: 10.1016/0167-2789(92)90242-F
[4]	RUDIN L I, OSHER S. Total variation based image restoration with free local constraints[C]// Proceedings of 1st International Conference on Image Processing Austin, IEEE, 1994: 31-35.
[5]	CHAN T F, SHEN J. Nontexture inpainting by curvature-driven diffusions[J]. Journal of Visual Communication and Image Representation, 2001, 12(4): 436-449. doi: 10.1006/jvci.2001.0487
[6]	CRIMINISI A, PÉREZ P, TOYAMA K. Region filling and object removal by exemplar-based image inpain-ting[J]. IEEE Transactions on Image Processing, 2004, 13(9): 1200-1212. doi: 10.1109/TIP.2004.833105
[7]	李张翼. 针对古织物图像的改进Criminisi修复算法[J]. 电脑知识与技术, 2016, 12(26):193-195.
	LI Zhangyi. Improved criminisi repair algorithm for ancient fabric images[J]. Computer Knowledge and Technology, 2016, 12(26):193-195.
[8]	朱耀麟, 李张翼, 武桐. 针对规则古织物纹理的图像修复[J]. 棉纺织技术, 2017, 45(10):9-12.
	ZHU Yaolin, LI Zhangyi, WU Tong. Image restoration for regular ancient fabric texture[J]. Cotton Textile Technology, 2017, 45(10):9-12.
[9]	韦秋菊, 孙晓婉, 徐平华, 等. 机织物局部组织复原与密度自动测定[J]. 现代纺织技术, 2022, 30(6): 102-109.
	WEI Qiuju, SUN Xiaowan, XU Pinghua, et al. Local weave restoration and automatic density measurement for fabrics[J]. Advanced Textile Technology, 2022, 30(6): 102-109.
[10]	BARNES C, SHECHTMAN E, FINKELSTEIN A, et al. PatchMatch: a randomized correspondence algorithm for structural image editing[C]// ACM SIGGRAPH Conference 2009. New Orleans: ACM. DOI:10.1145/1531326.1531330.
[11]	KORMAN S, AVIDAN S. Coherency sensitive hash-ing[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 38(6): 1099-1112. doi: 10.1109/TPAMI.2015.2477814
[12]	LECUN Y, BOSER B, DENKER J S, et al. Backpropagation applied to handwritten zip code recognition[J]. Neural computation, 1989, 1(4): 541-551. doi: 10.1162/neco.1989.1.4.541
[13]	RUMELHART D E, HINTON G E, WILLIAMS R J. Learning internal representations by error propaga-tion[J]. Parallel Distributed Processing, 1986, 1(1): 318-363.
[14]	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]// Advances in Neural Information Processing Systems 27 (NIPS 2014). Montreal: Curran Associates, Inc., 2014: 2672-2680.
[15]	SUVOROV R, LOGACHEVA E, MASHIKHIN A, et al. Resolution-robust large mask inpainting with fourier convolutions[C]// Proceedings of the IEEE/CVF Winter Conference On Applications Of Computer Vision. Waikoloa, IEEE, 2022: 2149-2159.
[16]	刘羿漩, 葛广英, 齐振岭, 等. 基于改进DCGAN的刺绣图像修复的研究[J]. 激光与光电子学进展, 2023, 60(20):1-20.
	LIU Yixuan, GE Guangying, QI Zhenling, et al. Research on embroidery image restoration based on improved DCGAN[J]. Laser & Optoelectronics Progress, 2023, 60(20):1-20.
[17]	TUKEY J W. Exploratory data analysis[M]. Reading: Addison-Wesley, 1971: 210-236.
[18]	TOMASI C, MANDUCHI R. Bilateral filtering for gray and color images[C]// Sixth International Conference on Computer Vision (IEEE Cat. No. 98CH36271). Bombay: IEEE, 1998: 839-846.
[19]	HE K, SUN J, TANG X. Guided image filtering[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 35(6): 1397-1409. doi: 10.1109/TPAMI.2012.213
[20]	XU L, YAN Q, XIA Y, et al. Structure extraction from texture via relative total variation[J]. ACM Transactions on Graphics (TOG), 2012, 31(6): 1-10.
[21]	CANNY J. A computational approach to edge detec-tion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1986 (6): 679-698.
[22]	LIU H, BI X, LU G, et al. Exemplar-based image inpainting with multi-resolution information and the graph cut technique[J]. IEEE Access, 2019, 7: 101641-101657. doi: 10.1109/Access.6287639
[23]	LE Meur O, EBDELLI M, GUILLEMOT C. Hierarchical super-resolution-based inpainting[J]. IEEE Transactions on Image Processing, 2013, 22(10): 3779-3790. doi: 10.1109/TIP.2013.2261308 pmid: 23661318
[24]	KWATRA V, SCHÖDL A, ESSA I, et al. Graphcut textures: image and video synthesis using graph cuts[J]. ACM Transactions on Graphics (TOG), 2003, 22(3): 277-286. doi: 10.1145/882262.882264
[25]	LEE J, LEE D K, PARK R H. Robust exemplar-based inpainting algorithm using region segmentation[J]. IEEE Transactions on Consumer Electronics, 2012, 58(2): 553-561. doi: 10.1109/TCE.2012.6227460
[26]	CHUNG B, YIM C. Hybrid error concealment method combining exemplar-based image inpainting and spatial interpolation[J]. Signal Processing: Image Communication, 2014, 29(10): 1121-1137. doi: 10.1016/j.image.2014.09.009
[27]	ZHANG L, ZHANG L, MOU X, et al. FSIM: A feature similarity index for image quality assessment[J]. IEEE Transactions on Image Processing, 2011, 20(8): 2378-2386. doi: 10.1109/TIP.2011.2109730 pmid: 21292594
[28]	CHEN L, JIANG F, ZHANG H, et al. Edge preservation ratio for image sharpness assessment[C]// 2016 12th World Congress on Intelligent Control and Automation (WCICA). Guilin:IEEE, 2016: 1377-1381.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed