基于CBAM残差块结合纹理采样器的水墨迁移算法

doi:10.16180/j.cnki.issn1007-7820.2025.06.010

电子科技 ›› 2025, Vol. 38 ›› Issue (6): 65-73.doi: 10.16180/j.cnki.issn1007-7820.2025.06.010

基于CBAM残差块结合纹理采样器的水墨迁移算法

刘雪莉¹, 杜洪波¹(), 袁雪丰¹, 朱立军²

1.沈阳工业大学理学院,辽宁沈阳 110870
2.北方民族大学信息与计算科学学院,宁夏银川 750021

收稿日期:2023-12-15 修回日期:2024-01-10 出版日期:2025-06-15 发布日期:2025-06-24
通讯作者: 杜洪波(1977-),男, E-mail:duhongbo@sut.edu.cn,副教授。研究方向:深度学习与图像处理。
作者简介:刘雪莉(1998-),女,硕士研究生。研究方向:风格迁移及图像处理。
基金资助:
国家自然科学基金(11861003);辽宁省教育厅高等学校基本科研项目(LJKZ0157)

Ink Transfer Algorithm Based on CBAM Residual Block Combined with Texture Sampler

LIU Xueli¹, DU Hongbo¹(), YUAN Xuefeng¹, ZHU Lijun²

1. School of Science,Shenyang University of Technology,Shenyang 110870,China
2. School of Information and Computing Science, Northern University for Nationalities, Yinchuan 750021, China

Received:2023-12-15 Revised:2024-01-10 Online:2025-06-15 Published:2025-06-24
Supported by:
National Natural Science Foundation of China(11861003);Basic Scientific Research Project of Higher Education Department of Liaoning(LJKZ0157)

摘要/Abstract

摘要：

针对水墨风格迁移算法出现的线条不清晰、纹理杂乱及颜色重建效果差等问题,文中提出了一种新型轻量级水墨风格迁移神经网络IWSGAN(Ink and Wash Style Generative Adversarial Networks)。使用CBAM(Convolutional Block Attention Module)残差块对特征图进行通道空间注意力计算,提高了算法对有效信息的采取率。结合水墨纹理显著性采样器ITSS(Ink-Texture-Saliency-Sampler)从训练数据中采样水墨显著性局部图像块,使迁移的水墨图片线条纹理的水墨质感更明显。采用5种不同的损失函数约束内容图像和生成图像的高层语义,以促进风格特征的一致性。实验结果表明,IWSGAN生成图像保留了更多的内容特征,线条的纹理细节更形象,色彩重建表现较好,图像质量得到了显著提升。

关键词: 风格迁移, 生成对抗网络, CBAM残差块, 注意力机制, ITSS, 局部对抗, 水墨迁移, 消融实验

Abstract:

In view of the problems of ink style transfer algorithm, such as unclear lines, chaotic textures and poor color reconstruction, a new lightweight ink style transfer neural network IWSGAN(Ink and Wash Style Generative Adversarial Networks) is proposed in this study. The CBAM(Convolutional Block Attention Module) residual block is used to calculate the channel spatial attention of the feature map, which improves the algorithm's adoption rate of effective information. In combination with the ink texture salience sampler ITSS(Ink-Texture-Saliency-Sampler), the local image blocks of ink painting salience are sampled from the training data to make the line texture of the transferred ink picture more ink texture. Five different loss functions are used to constrain the high-level semantics of the content image and the generated image to facilitate consistency in style characteristics. The experimental results show that the image generated by IWSGAN retains more content features, the texture details of the lines are more vivid, the color reconstruction performance is excellent, and the image quality is significantly improved.

Key words: style transfer, generative adversarial network, CBAM residual block, attention mechanism, ITSS, local adversarial, ink transfer, ablation experiment

中图分类号:

TP391.41

刘雪莉, 杜洪波, 袁雪丰, 朱立军. 基于CBAM残差块结合纹理采样器的水墨迁移算法[J]. 电子科技, 2025, 38(6): 65-73.

LIU Xueli, DU Hongbo, YUAN Xuefeng, ZHU Lijun. Ink Transfer Algorithm Based on CBAM Residual Block Combined with Texture Sampler[J]. Electronic Science and Technology, 2025, 38(6): 65-73.

图/表 15

图1

图2

图3

图4

图5

图6

图7

图8

图9

图10

表1

表2

表3

图11

图12

参考文献 25

[1]	Schroeder B, Tripathi S. Structured query-based image retrieval using scene graphs[C]. Seattle: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops,2020:178-179.
[2]	Yoon S, Kang W Y, Jeon S, et al. Image-to-image retrieval by learning similarity between scene graphs[C]. Vancouver: Proceedings of the AAAI Conference on Artificial Intelligence,2021:10718-10726.
[3]	程俊华, 曾国辉, 刘瑾. 基于深度学习的复杂背景图像分类方法研究[J]. 电子科技, 2020, 33(12):59-66.
	Cheng Junhua, Zeng Guohui, Liu Jin. Research on complex background image classification method based on deep learning[J]. Electronic Science and Technology, 2020, 33(12):59-66.
[4]	Goodfellow I, Pouget Abadie J, Mirza M, et al. Generative adversarial nets[C]. Montreal: Proceedings of International Conference on Neural Information Processing Systems, 2014:1-9
[5]	Pang Y, Lin J, Qin T, et al. Image-to-image translation: Methods and applications[J]. IEEE Transactions on Multimedia, 2021, 24(1):3859-3881.
[6]	李乐阳, 佟国香, 赵迎志, 等. 基于生成对抗网络的文本生成图像研究综述[J]. 电子科技, 2023, 36(10):39-55.
	Li Yueyang, Tong Guoxiang, Zhao Yingzhi, et al. A survey of text-to-image synthesis based on generative adversarial network[J]. Electronic Science and Technology, 2023, 36(10):39-55.
[7]	Liang J, Cao J, Sun G, et al. Swinir:Image restoration using swin transformer[C]. Montreal: Proceedings of the IEEE/CVF International Conference on Computer Vision,2021:1833-1844.
[8]	Deng Y, Tang F, Dong W, et al. Stytr2:Image style transfer with transformers[C]. New Orleans: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2022:11326-11336.
[9]	Zhang H, Sindagi V, Patel V M. Image deraining using a conditional generative adversarial network[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 30(11):3943-3956.
[10]	Kim D D, Shahid M T, Kim Y, et al. Generating pedestrian training dataset using DCGAN[C]. Chengdu: Proceedings of the Third International Conference on Advances in Image Processing,2019:1-4.
[11]	Isola P, Zhu J Y, Zhou T, et al. Image-to-image translation with conditional adversarial networks[C]. Honolulu: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2017:1125-1134.
[12]	Huang X, Belongie S. Arbitrary style transfer in real-time with adaptive instance normalization[C]. Zhangjiajie: Proceedings of the IEEE International Conference on Computer Vision,2017:1501-1510.
[13]	Zhu J Y, Park T, Isola P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]. Venice: Proceedings of the IEEE International Conference on Computer Vision,2017:2223-2232.
[14]	Xu W, Long C, Wang R, et al. Drb-gan:A dynamic resblock generative adversarial network for artistic style transfer[C]. Montreal: Proceedings of the IEEE/CVF International Conference on Computer Vision,2021:6383-6392.
[15]	Lin C T, Huang S W, Wu Y Y, et al. GAN-based day-to-night image style transfer for nighttime vehicle detection[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 22(2):951-963.
[16]	Zhou Y, Yu K, Wang M, et al. Speckle noise reduction for OCT images based on image style transfer and conditional GAN[J]. IEEE Journal of Biomedical and Health Informatics, 2021, 26(1):139-150.
[17]	Shu Y, Yi R, Xia M, et al. Gan-based multi-style photo cartoonization[J]. IEEE Transactions on Visualization and computer graphics, 2021, 28(10):3376-3390.
[18]	Gu J, Shen Y, Zhou B. Image processing using multi-code gan prior[C]. Seattle: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2020:3012-3021.
[19]	He B, Gao F, Ma D, et al. Chipgan:A generative adversarial network for chinese ink wash painting style transfer[C]. Seoul: Proceedings of the Twenty-sixth ACM International Conference on Multimedia,2018:1172-1180.
[20]	Liu S, Lin T, He D, et al. Adaattn:Revisit attention mechanism in arbitrary neural style transfer[C]. Montreal: Proceedings of the IEEE/CVF International Conference on Computer Vision,2021:6649-6658.
[21]	Wu X, Hu Z, Sheng L, et al. Styleformer:Real-time arbitrary style transfer via parametric style composition[C]. Montreal: Proceedings of the IEEE/CVF International Conference on Computer Vision,2021:14618-14627.
[22]	Lin T, Ma Z, Li F, et al. Drafting and revision:Laplacian pyramid network for fast high-quality artistic style transfer[C]. Online: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2021:5141-5150.
[23]	An J, Huang S, Song Y, et al. Artflow:Unbiased image style transfer via reversible neural flows[C]. Nashville: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,2021:862-871.
[24]	Chen J, Liu G, Chen X. AnimeGAN:A novel lightweight GAN for photo animation[C]. Guangzhou: International Symposium on Intelligence Computation and Applications, Springer,Singapore,2020: 242-256.
[25]	Woo S, Park J, Lee J Y, et al. Cbam:Convolutional block attention module[C]. Munich: Proceedings of the European Conference on Computer Vision,2018: 3-19.

数据集	AdaIN	CycleGAN	ChipGAN	AdaAttN	IWSGAN
Scenery	202.64	182.53	190.47	161.42	156.72
Hourse	177.61	163.52	111.46	125.85	102.47

数据集	AdaIN	CycleGAN		ChipGAN		AdaAttN		IWSGAN
Scenery	15.32±0.68		12.68±0.65		9.49±0.35		7.37±0.24		7.14±0.62
Hourse	17.73±0.44		13.52±0.79		10.25±0.28		6.94±0.31		5.84±0.47

数据集	AdaIN		CycleGAN		ChipGAN		AdaAttN			IWSGAN
Scenery		4.957		4.177		5.602		6.457	6.826
Hourse		5.236		4.341		4.936		5.482	5.697

基于CBAM残差块结合纹理采样器的水墨迁移算法

Ink Transfer Algorithm Based on CBAM Residual Block Combined with Texture Sampler

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 15

参考文献 25

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	唐玉良, 张轩雄. 基于深度学习的车位检测方[J]. 电子科技, 2025, 38(6): 23-29.
[2]	陈清江, 杨双. 基于多层特征增强的双分支图像去雾算法[J]. 电子科技, 2025, 38(6): 30-38.
[3]	陈漫漫, 于莲芝. 结合共注意网络的深度BiGRU和DPCS情感分析模型[J]. 电子科技, 2025, 38(5): 22-30.
[4]	胡锦磊, 田恩刚, 瞿枫. 融合通道全局注意力机制的双路径实时语义分割[J]. 电子科技, 2025, 38(5): 83-88.
[5]	傅鹏, 宋晓霞. 基于CycleGAN的地震数据去噪方法[J]. 电子科技, 2025, 38(4): 25-30.
[6]	王勇, 杨义龙, 范晓晖, 周雷, 孔祥勇. 基于迁移学习和改进EfficientNet-B0的脑肿瘤分类算法[J]. 电子科技, 2025, 38(4): 46-51.
[7]	周凯, 于莲芝. 融合注意力机制的弱监督语义分割自激活方法[J]. 电子科技, 2025, 38(4): 80-86.
[8]	陈宇洋, 李峰. 融合CNN与Transformer的视网膜OCT图像积液分割方法[J]. 电子科技, 2025, 38(3): 47-59.
[9]	郑方亮, 王延年, 廉继红, 阮佩. 基于条件先验Swin Transformer的人脸图像超分辨重建[J]. 电子科技, 2025, 38(2): 35-41.
[10]	赖颖, 巨志勇, 叶雨新. 基于改进YOLOv4的车辆检测算法[J]. 电子科技, 2025, 38(1): 81-87.
[11]	蒯新晨, 李烨. 多种及多尺度注意力混合的图像超分辨率重建[J]. 电子科技, 2024, 37(9): 34-42.
[12]	何幸, 黄永明, 朱勇. 基于改进YOLOv5的路面坑洼检测方法[J]. 电子科技, 2024, 37(7): 53-59.
[13]	王圣雄, 刘瑞安, 燕达. 基于改进生成对抗网络的图像风格迁移算法[J]. 电子科技, 2024, 37(6): 36-43.
[14]	叶雨新, 巨志勇, 赖颖. 融合感受野增强和注意力机制的交通标志检测算法[J]. 电子科技, 2024, 37(6): 8-16.
[15]	朱子豪, 宋燕. 融合注意力和胶囊池化的轻量型胶囊网络[J]. 电子科技, 2024, 37(5): 1-8.