|
|
|
| Reversible Flow Network for Style Transfer Based on Frequency-Domain Enhanced Adaptive Channel Attention and Feature Pyramid Fusion |
| GE Bin1,2, SHAO Ziyi1,2, ZHENG Junshuai1, XIA Chenxing1, GUAN Junming3, XU Tao1 |
1. State Key Laboratory of Digital Intelligent Technology for Unmanned Coal Mining, Anhui University of Science and Technology, Huainan 232001;
2. College of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001;
3. School of Information Engineering, HuangShan University, Huangshan 245041 |
|
|
|
|
Abstract To address the issues of content distortion, artifact appearance, and insufficient utilization of frequency-domain characteristics in style transfer, a reversible flow network for style transfer based on frequency-domain enhanced adaptive channel attention and feature pyramid fusion(FECANet) is proposed. Based on the pre-trained VGG19 architecture, a reversible flow network is designed to reduce feature loss and ensure the integrity of content structure by leveraging its unbiased feature transfer mechanism. A frequency-domain enhanced adaptive channel attention module is developed to analyze the frequency-domain distribution of style images, and accurate correlations between content and style features are established to improve the stylization effect. Additionally, a feature pyramid fusion scheme is designed to align global style with local textures, enhancing the coordination of transfer results. Experiments on MS-COCO and WikiArt datasets show that FECANet effectively balances style transfer and content preservation, and it shows superior performance in content structure integrity, stylization effect and computational efficiency.
|
|
Received: 04 December 2025
|
|
|
| Fund:General Program of National Natural Science Foundation of China(No.52374155), Graduate Innovation Fund of Anhui University of Science and Technology(No.2024cx2105) |
|
Corresponding Authors:
ZHENG Junshuai, Ph.D., lecturer. His research interests include computer vision, deepfake detection, and contrastive learning.
|
About author:: GE Bin, Ph.D., professor. His research interests include computer vision, pattern re-cognition, and information security.
SHAO Ziyi, Master student. His research interests include computer vision and deep learning.
XIA Chenxing, Ph.D., associate pro-fessor. His research interests include compu-ter vision, pattern recognition, and machine learning.
GUAN Junming, Ph.D., associate pro-fessor. His research interests include compu-ter vision and pattern recognition.
XU Tao, Master student. His research interests include computer vision and deep learning. |
|
|
|
[1] WANG H, LI Y J, WANG Y H, et al. Collaborative Distillation for Ultra-Resolution Universal Style Transfer // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2020: 1857-1866.
[2] GATYS L A, ECKER A S, BETHGE M.Image Style Transfer Using Convolutional Neural Networks // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2016: 2414-2423.
[3] JOHNSON J, ALAHI A, LI F F.Perceptual Losses for Real-Time Style Transfer and Super-Resolution // Proc of the 14th European Conference on Computer Vision. Berlin, Germany: Springer, 2016: 694-711.
[4] SIMONYAN K, ZISSERMAN A.Very Deep Convolutional Networks for Large-Scale Image Recognition. Computer Science, 2014, 41(9): 103951-103958.
[5] 林星,陈昭炯,叶东毅.主体结构保持的增强型图像风格传递方法.模式识别与人工智能, 2018, 31(11): 997-1007.
(LIN X, CHEN Z J, YE D Y.Enhanced Image Style Transferring Method with Primary Structure Maintained. Pattern Recognition and Artificial Intelligence, 2018, 31(11): 997-1007.)
[6] HUANG X, BELONGIE S.Arbitrary Style Transfer in Real-Time with Adaptive Instance Normalization // Proc of the IEEE International Conference on Computer Vision. Washington, USA: IEEE, 2017: 1510-1519.
[7] LI Y J, FANG C, YANG J M, et al.Universal Style Transfer via Feature Transforms // Proc of the 31st International Conference on Neural Information Processing Systems. Cambridge,USA: MIT Press, 2017: 385-395.
[8] PARK D Y, LEE K H.Arbitrary Style Transfer with Style-Attentional Networks // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2019: 5873-5881.
[9] LIU S H, LIN T W, HE D L, et al. AdaAttN: Revisit Attention Mechanism in Arbitrary Neural Style Transfer // Proc of the IEEE/CVF International Conference on Computer Vision. Washington, USA: IEEE, 2021: 6629-6638.
[10] PENG H Y, QIAN W H, CAO J D,et al. Arbitrary Style Transfer Based on Attention and Covariance-Matching. Computers & Gra-phics, 2023, 116: 298-307.
[11] YANG G M, ZHANG S C, FANG X J, et al. Pyramid Style-Attentional Network for Arbitrary Style Transfer. Multimedia Tools and Applications, 2024, 83(5): 13483-13502.
[12] AN J, XIONG H Y, HUAN J, et al. Ultrafast Photorealistic Style Transfer via Neural Architecture Search. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 10443-10450.
[13] WU Z J, ZHU Z, DU J P,et al. CCPL: Contrastive Coherence Preserving Loss for Versatile Style Transfer // Proc of the 17th European Conference on Computer Vision. Berlin, Germany: Sprin- ger, 2022: 189-206.
[14] YE Z X, HUANG H J, WANG X T, et al. StyleMaster: Stylize Your Video with Artistic Generation and Translation // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2025: 2630-2640.
[15] BAI Z Y, XU H L, DING Q C, et al. MCAFNet: Multi-scale Channel Attention Fusion Network for Arbitrary Style Transfer. IEEE Transactions on Instrumentation and Measurement, 2025, 74. DOI: 10.1109/TIM.2025.3561400.
[16] MA P C, YANG X P, LI Y S, et al. SCFlow: Implicitly Learning Style and Content Disentanglement with Flow Models[C/OL].[2025-11-10]. https://arxiv.org/pdf/2508.03402v1.
[17] DINH L, KRUEGER D W, BENGIO Y.NICE: Non-linear Independent Components Estimation[C/OL]. [2025-11-10].https://arxiv.org/pdf/1410.8516.
[18] KINGMA D P, DHARIWAL P.Glow: Generative Flow with Invertible 1×1 Convolutions[C/OL]. [2025-11-10].https://arxiv.org/pdf/1807.03039.
[19] AN J, HUANG S Y, SONG Y B, et al. ArtFlow: Unbiased Image Style Transfer via Reversible Neural Flows // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2021: 862-871.
[20] LI H A, WANG L Y, LIU J.A Review of Deep Learning-Based Image Style Transfer Research. The Imaging Science Journal, 2025, 73(4): 504-526.
[21] SU N, WANG J T, PAN Y.Multi-scale Universal Style-Transfer Network Based on Diffusion Model. Algorithms, 2025, 18(8), DOI: 10.3390/a18080481.
[22] WEN L F, GAO C Y, ZOU C Q.CAP-VSTNet: Content Affinity Preserved Versatile Style Transfer // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2023: 18300-18309.
[23] FENG W C, FENG W Q, HUANG D W, et al. StyleBrush: Style Extraction and Transfer from a Single Image. Computer Science, 2024, 51(8): 104862-104870.
[24] HU Z S, GE B, XIA C X, et al. Flow Style-Aware Network for Arbitrary Style Transfer. Computers & Graphics, 2024, 124. DOI: 10.1016/j.cag.2024.104098.
[25] QIN Z Q, ZHANG P Y, WU F, et al. FcaNet: Frequency Cha-nnel Attention Networks // Proc of the IEEE/CVF International Conference on Computer Vision. Washington, USA: IEEE, 2021: 763-772.
[26] CHEN D D, YUAN L, LIAO J, et al. StyleBank: An Explicit Representation for Neural Image Style Transfer // Proc of the IEEE Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2017: 2770-2779.
[27] KWON J, KIM S, LIN Y W, et al. AesFA: An Aesthetic Feature-Aware Arbitrary Neural Style Transfer.Proceedings of the AAAI Conference on Artificial Intelligence, 2024, 38(12): 13310-13319.
[28] DONG Y S, LIU S T, LI Y X, et al. Aesthetic-Aware Adversarial Learning Network for Artistic Style Transfer. Neurocomputing, 2025, 646. DOI: 10.1016/j.neucom.2025.130431.
[29] XU R J, XI W J, WANG X D, et al. StyleSSP: Sampling StartPoint Enhancement for Training-Free Diffusion-Based Method for Style Transfer // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2025: 18260-18269. |
|
|
|
2026, Vol. 39 
