| 93 | 0 | 7 |
| 下载次数 | 被引频次 | 阅读次数 |
中国传统画作为宝贵的文化遗产，历经时间沉淀以及各种自然因素的影响，常出现开裂、破损和褪色等问题。尽管一些深度学习框架在自然图像修复领域取得了显著进展，但其大多过度依赖卷积权重共享和平移不变性，在处理布局复杂、结构抽象的绘画图像时，难以捕捉其独特的空间特性。针对此问题，提出一种孪生级联空间滤波(twin cascade spatial filtering, TCSF)预测方法用于中国传统画的修复。TCSF采用层级解码策略，从多尺度解析绘画图像的层次特征，并级联空间滤波预测方法得到修复核，从而由粗到细地复原缺失区域的像素。为了在特征信息匮乏的区域精确地复原缺失的结构和笔触信息，进一步引入空间编码机制。通过对滤波特征图空间编码得到坐标矩阵，并在滤波预测过程中注入坐标信息编码，用于缺失像素点恢复时提供空间信息参照，进而提升修复结果的精确度与视觉效果。实验中，选取了具有代表性的中国传统画图像进行训练，并增加壁画数据集和Places数据集测试模型的泛化性能。与现有工作使用掩码不同，该研究在实验中提取部分真实绘画图像的破损掩码，以更逼真地模拟破损情况。定性和定量实验结果表明，该方法在中国传统画恢复任务中取得了较好的修复结果，为数字艺术修复和文化遗产保护提供了有益的启示。
Abstract：Traditional Chinese paintings are invaluable cultural legacies, but they often suffer from issues such as cracking, damage, and fading due to the effects of time and various natural factors. While some deep learning frameworks have made significant progress in natural image restoration, they tend to rely heavily on convolutional weight sharing and translational invariance. This reliance may limit their ability to fully capture the unique spatial characteristics of paintings with intricate layouts and abstract structural information. To address this issue, this paper proposes a Twin Cascade Spatial Filtering(TCSF) prediction method for the restoration of traditional Chinese paintings. The TCSF adopts a hierarchical decoding strategy that analyzes the hierarchical features of painting images across multiple scales. It cascades a spatial filtering prediction approach to obtain restoration kernels, restoring missing region pixels from coarse to fine detail. Furthermore, in order to precisely restore the missing structural and brushstroke information in areas where feature information is sparse, this paper introduces a spatial encoding mechanism. By spatially encoding the filter feature maps into coordinate matrices and infusing coordinate information encoding into the filtering prediction process, this paper provides spatial reference information for the recovery of missing pixels, thereby enhancing the accuracy and visual quality of the restoration outcomes. In the experiments, the model was trained using representative images of traditional Chinese paintings, and the mural datasets and Places datasets were added to test the model's generalization ability. In contrast to existing work that utilizes synthetic masks, this paper extracted actual damage masks from real painting images in order to more realistically simulate damage scenarios. The qualitative and quantitative experimental results demonstrate that the proposed method achieves favorable restoration results in traditional Chinese painting recovery tasks and provides useful insights for digital art restoration and cultural heritage protection.
[1] LI J,WANG N,ZHANG L,et al.Recurrent feature reasoning for image inpainting[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle:IEEE,2020:7760-7768.
[2] 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.
[3] WAN Z,ZHANG J,CHEN D,et al.High-fidelity pluralistic image completion with transformers[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Montreal:IEEE,2021:4692-4701.
[4] LIAO L,XIAO J,WANG Z,et al.Guidance and evaluation:Semantic-aware image inpainting for mixed scenes[C]//Computer Vision-ECCV 2020:16th European Conference.Cham:Springer,2020:683-700.
[5] LI L,ZOU Q,ZHANG F,et al.Line drawing guided progressive inpainting of mural damages[EB/OL].(2022-11-12) [2024-05-15].https://arxiv.org/abs/2211.06649.
[6] ZHU J Y,PARK T,ISOLA P,et al.Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE International Conference on Computer Vision.Venice:IEEE,2017:2223-2232.
[7] 曹建芳，靳梦燕，李朝霞，等.基于循环生成对抗网络的壁画色彩修复算法[J].山东科技大学学报(自然科学版),2023,42(4):101-112.CAO J F,JIN M Y,LI Z X,et al.Mural color restoration algorithm based on cyclic generative adversarial network[J].Journal of Shandong University of Science and Technology (Natural Science),2023,42(4):101-112.
[8] CHI L,JIANG B,MU Y.Fast fourier convolution[J].Advances in Neural Information Processing Systems,2020,33:4479-4488.
[9] YU J,LIN Z,YANG J,et al.Free-form image inpainting with gated convolution[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Seoul:IEEE,2019:4471-4480.
[10] 胡升，薛涛，季虹.多尺度信息融合的生成对抗网络壁画修复[J].国外电子测量技术，2024,43(4):30-38.HU S,XUE T,JI H.Muralinpainting algorithm for generative adversarial network with multi-scale information fusion[J].Foreign Electronic Measurement Technology,2024,43(4):30-38.
[11] 赵磊，吉柏言，邢卫，等.基于多路编码器和双重注意力的古画修复算法[J].计算机研究与发展，2023,60(12):2814-2831.ZHAO L,JI B Y,XING W,et al.Ancient painting inpaintingalgorithm based on multi-channel encoder and dualattention[J].Joural of Computer Research and Development,2023,60(12):2814-2831.
[12] BARNES C,SHECHTMAN E,FINKELSTEIN A,et al.PatchMatch:A randomized correspondence algorithm for structural image editing[J].ACM Trans Graph,2009,28(3):1-11.
[13] CHO T S,BUTMAN M,AVIDAN S,et al.The patch transform and its applications to image editing[C]//2008 IEEE Conference on Computer Vision and Pattern Recognition.Anchorage:IEEE,2008:1-8.
[14] KWATRA V,ESSA I,BOBICK A,et al.Texture optimization for example-based synthesis[J].ACM Siggraph 2005 Papers,2005,24(3):795-802.
[15] DARABI S,SHECHTMAN E,BARNES C,et al.Image melding:Combining inconsistent images using patch-based synthesis[J].ACM Transactions on graphics (TOG),2012,31(4):1-10.
[16] ZOME T.Learning how to inpaint from global image statistics[C]//Proceedings Ninth IEEE International Conference on Computer Vision.Nice:IEEE,2003:305-312.
[17] BALLESTER C,BERTALMIO M,CASELLES V,et al.Filling-in by joint interpolation of vector fields and gray levels[J].IEEE Transactions on Image Processing,2001,10(8):1200-1211.
[18] PATHAK D,KRAHENBUHL P,DONAHUE J,et al.Context encoders:Feature learning by inpainting[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:2536-2544.
[19] LI W,LIN Z,ZHOU K,et al.Mat:Mask-aware transformer for large hole image inpainting[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:10758-10768.
[20] DONG Q,CAO C,FU Y.Incremental transformer structure enhanced image inpainting with masking positional encoding[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:11358-11368.
[21] KO K,KIM C S.Continuously masked transformer for image inpainting[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Paris:IEEE,2023:13169-13178.
[22] CHEN B W,LIU T J,LIU K H.Lightweight image inpainting by stripe window transformer with joint attention to CNN[C]//2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing (MLSP).Rome:IEEE,2023:1-6.
[23] LIU H,JIANG B,XIAO Y,et al.Coherent semantic attention for image inpainting[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Seoul:IEEE,2019:4170-4179.
[24] YAN Z,LI X,LI M,et al.Shift-net:Image inpainting via deep feature rearrangement[C]//Proceedings of the European Conference on Computer Vision (ECCV).Cham:Springer,2018:1-17.
[25] WANG N,LI J,ZHANG L,et al.MUSICAL:Multi-scale image contextual attention learning for inpainting[C]//Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence (IJCAI).Macao:IJCAI,2019:3748-3754.
[26] YU J,LIN Z,YANG J,et al.Generative image inpainting with contextual attention[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:5505-5514.
[27] ZHENG H,LIN Z,LU J,et al.Image inpainting with cascaded modulation gan and object-aware training[C]//European Conference on Computer Vision.Cham:Springer,2022:277-296.
[28] 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.
[29] IIZUKA S,SIMO-SERRA E,ISHIKAWA H.Globally and locally consistent image completion[J].ACM Transactions on Graphics (ToG),2017,36(4):1-14.
[30] GUO X,YANG H,HUANG D.Image inpainting via conditional texture and structure dual generation[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Montreal:IEEE,2021:14134-14143.
[31] ZENG Y,LIN Z,LU H,et al.Cr-fill:Generative image inpainting with auxiliary contextual reconstruction[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.Montreal:IEEE,2021:14164-14173.
[32] SONG Y,YANG C,SHEN Y,et al.Spg-net:Segmentation prediction and guidance network for image inpainting[EB/OL].(2018-08-06) [2024-05-15].https://arxiv.org/abs/1805.03356.
[33] ZHENG C,CHAM T J,CAI J,et al.Bridging global context interactions for high-fidelity image completion[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:11512-11522.
[34] DENG Y,HUI S,ZHOU S,et al.T-former:An efficient transformer for image inpainting[C]//Proceedings of the 30th ACM International Conference on Multimedia.Lisboa:ACM,2022:6559-6568.
[35] DENG Y,HUI S,MENG R,et al.Hourglass attention network for image inpainting[C]//European Conference on Computer Vision.Cham:Springer,2022:483-501.
[36] MILDENHALL B,BARRON J T,CHEN J,et al.Burst denoising with kernel prediction networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:2502-2510.
[37] XIE C,TIAN X,JIANG R,et al.Dilated kernel prediction network for single-image denoising[J].Journal of Electronic Imaging,2021,30(2):1-15.
[38] CARBAJAL G,VITORIA P,LEZAMA J,et al.Blind motion deblurring with pixel-wise kernel estimation via kernel prediction networks[J].IEEE Transactions on Computational Imaging,2023,9:928-943.
[39] GUO Q,SUN J,JUEFEI-XU F,et al.Efficientderain:Learning pixel-wise dilation filtering for high-efficiency single-image deraining[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Palo Alto:AAAI Press,2021:1487-1495.
[40] GUO Q,LI X,XU F,et al.Jpgnet:Joint predictive filtering and generative network for image inpainting[C]//Proceedings of the 29th ACM International Conference on Multimedia.New York:ACM,2021:386-394.
[41] LI X,GUO Q,LIN D,et al.Misf:Multi-level interactive siamese filtering for high-fidelity image inpainting[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans:IEEE,2022:1869-1878.
[42] LIU R,LEHMAN J,MOLINO P,et al.An intriguing failing of convolutional neural networks and the coordconv solution[J].Advances in Neural Information Processing Systems,2018,31:9605-9616.
[43] JOHNSON J,ALAHI A,FEI-FEI L.Perceptual losses for real-time style transfer and super-resolution[C]//Computer Vision-ECCV 2016:14th European Conference.Cham:Springer,2016:694-711.
[44] GOODFELLOW I,POUGET-ABADIE J,MIRZA M,et al.Generative adversarial nets[J].Advances in Neural Information Processing Systems,2014,27:2672-2680.
[45] GATYS L A,ECKER A S,BETHGE M.Image style transfer using convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas:IEEE,2016:2414-2423.
[46] ZHOU B,LAPEDRIZA A,KHOSLA A,et al.Places:A 10 million image database for scene recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,40(6):1452-1464.
[47] SIMONYAN K.Very deep convolutional networks for large-scale image recognition[EB/OL].(2015-04-10) [2024-05-15].https://arxiv.org/abs/1409.1556.
[48] YI Z,TANG Q,AZIZI S,et al.Contextual residual aggregation for ultra high-resolution image inpainting[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle:IEEE,2020:7508-7517.
[49] LIU G,REDA F A,SHIH K J,et al.Image inpainting for irregular holes using partial convolutions[C]//Proceedings of the European Conference on Computer Vision (ECCV).Cham:Springer,2018:85-100.
[50] ZHANG R,ISOLA P,EFROS A A,et al.The unreasonable effectiveness of deep features as a perceptual metric[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City:IEEE,2018:586-595.
[51] PASZKE A,GROSS S,MASSA F,et al.Pytorch:An imperative style,high-performance deep learning library[J].Advances in Neural Information Processing Systems,2019,32:8026-8037.
[52] LOSHCHILOV I.Decoupled weight decay regularization[EB/OL].(2019-01-04) [2024-05-15].https://arxiv.org/abs/1711.05101.
[53] ZENG Y,FU J,CHAO H,et al.Aggregated contextual transformations for high-resolution image inpainting[J].IEEE Transactions on Visualization and Computer Graphics,2022,29(7):3266-3280.
[54] LIU W,CUN X,PUN C M,et al.Coordfill:Efficient high-resolution image inpainting via parameterized coordinate querying[C]//Proceedings of the AAAI Conference on Artificial Intelligence.Washington:AAAI Press,2023,37(2):1746-1754.
基本信息:
DOI：10.16152/j.cnki.xdxbzr.2025-01-013
中图分类号:K879.4;J212;TP391.41
引用信息:
[1]薛文喆,董兴宇,胡琦瑶等.基于孪生级联空间滤波的中国传统画修复[J].西北大学学报(自然科学版),2025,55(01):150-167.DOI:10.16152/j.cnki.xdxbzr.2025-01-013.
基金信息:
国家自然科学基金(62471390,62306237); 陕西省重点研发计划(2024GX-YBXM-149); 西北大学研究生创新项目(CX2024204、CX2024206)