计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (5): 237-246.doi: 10.11896/jsjkx.250400097

• 计算机图形学 & 多媒体 • 上一篇    下一篇

基于自注意力隐式特征编解码的图像连续超分辨率算法

陈柏颖, 石颉   

  1. 苏州科技大学电子与信息工程学院 江苏 苏州 215000
  • 收稿日期:2025-04-21 修回日期:2025-07-02 发布日期:2026-05-08
  • 通讯作者: 石颉(17751455752@163.com)
  • 作者简介:(1079510262@qq.com)
  • 基金资助:
    国家自然科学基金面上项目(52477114)

Continuous Image Super-resolution Based on Self-attention Implicit Feature Encoding andDecoding

CHEN Boying, SHI Jie   

  1. School of Electronics and Information Engineering, Suzhou University of Science and Technology, Suzhou, Jiangsu 215000, China
  • Received:2025-04-21 Revised:2025-07-02 Online:2026-05-08
  • About author:CHEN Boying,born in 1999,postgra-duate.His main research interests include deep learning and image super-resolution algorithm.
    SHI Jie,born in 1978,Ph.D,professor.His main research interests include deep learning,artificial intelligence,intelligent perception of power equipment.
  • Supported by:
    General Projects of the National Natural Science Foundation of China(52477114).

PDF (PC)

243

摘要： 针对卷积神经网络在图像超分辨率重建中只能处理固定分辨率,难以实现任意分辨率大小的连续超分辨率重建的问题,提出了一种基于自注意力隐式特征编解码的连续倍率图像超分辨重建算法。首先,通过基于局部-全局自注意力联合建模的特征编码网络,实现对低分辨率图像到高维特征的特征映射;其次,利用局部隐式特征强化编码器来有效聚合局部邻域特征,通过自注意力机制来增强特征邻域数据之间的关联性;最后,采用粗粒度与细粒度像素特征协同关联的多粒度隐式特征解码器,将图像坐标和坐标邻域的深度特征输入多层感知机来预测高分辨率坐标的像素值。实验结果表明,与目前较好的图像重建算法模型相比,所提方法取得了较好的超分辨率重建效果,具有优越性和有效性。

关键词: 连续超分辨率重建, 神经隐式表示, 局部隐式特征, 注意力机制

Abstract: In order to solve the problem that the convolutional neural network can only deal with fixed resolution in image super-resolution reconstruction,and it is difficult to realize continuous super-resolution reconstruction with arbitrary resolution,this paper proposes an algorithm for super-resolution reconstruction of continuous-rate images based on self-attention implicit feature encoding and decoding.Firstly,the feature encoding network based on local-global self-attention modeling is used to realize the feature mapping from low-resolution image to high-dimensional feature.Then local implicit feature enhancement encoders are used to effectively aggregate local neighborhood features,and self-attention mechanism is used to enhance the correlation between feature neighborhood data.Finally a multi-granularity implicit feature decoder is used to predict the pixel values of high-resolution coordinates by inputting the image coordinates and the depth features of the adjacent coordinate multilayer perceptron.Experimental results show that,compared with the current image reconstruction algorithm,the proposed method achieves better super-resolution reconstruction results,which proves the superiority and effectiveness of the method.

Key words: Continuous super-resolution reconstruction, Neural implicit representation, Local implicit features, Attention mechanism

中图分类号: 

  • TP391
[1]CAO S M,WANG X H,WANG H K,et al.MSV-Net:Visual Super-Resolution Reconstruction for Scientific Simulated Data of Mixed Surface-Volume[J].Journal of Frontiers of Computer Science and Technology,2023,17(6):1321-1328.
[2]LYU J,XU P C.Image Super-resolution Reconstruction Algorithm Based on Multi-scale Adaptive Upsampling[J].Journal of Frontiers of Computer Science and Technology,2023,17(4):879-891.
[3]MAHAPATRA D,BOZORGTABAR B,GARNAVI R.Imagesuper-resolution using progressive generative adversarial networks for medical image analysis[J].Computerized Medical Imaging and Graphics,2019,71:30-39.
[4]UMIRZAKOVA S,AHMAD S,KHAN L U,et al.Medicalimage super-resolution for smart healthcare applications:A comprehensive survey[J].Information Fusion,2024,103:102075.
[5]QIU D,CHENG Y,WANG X.Medical image super-resolution reconstruction algorithms based on deep learning:A survey[J].Computer Methods and Programs in Biomedicine,2023,238:107590.
[6]XIAO Y,YUAN Q,JIANG K,et al.TTST:A top-k token selective transformer for remote sensing image super-resolution[J].IEEE Transactions on Image Processing,2024,33:738-752.
[7]HAN X L,ZHANG H,SUN W D.Spectral super-resolutionusing optimized dictionary learning via spectral library and its effects on classification[J].National Remote Sensing Bulletin,2023,27(11):2530-2540.
[8]LIM B,SON S,KIM H,et al.Enhanced deep residual networks for single image super-resolution[C]//IEEE Computer Society Conference on Computer Vision & Pattern Recognition.2017:1132-1140.
[9]ZHANG Y,LI K,LI K,et al.Image super-resolution using very deep residual channel attention networks[C]//Proceedings of the European Conference on Computer Vision(ECCV).2018:286-301.
[10]ZHANG Y,TIAN Y,KONG Y,et al.Residual dense network for image super-resolution[C]//IEEE Computer Society Conference on Computer Vision & Pattern Recognition.2018:2472-2481.
[11]MILDENHALL B,SRINIVASAN P P,TANCIK M,et al.Nerf:Representing scenes as neural radiance fields for view synthesis[J].Communications of the ACM,2021,65(1):99-106.
[12]CHIBANE J,ALLDIECK T,PONS-MOLL G.Implicit functions in feature space for 3d shape reconstruction and completion[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:6970-6981.
[13]CHEN Y B,LIU S F,WANG X L.Learning continuous image representation with local implicit image function[C]//Procee-dings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2021.
[14]XU X Q,WANG Z Y,SHI H.Ultrasr:Spatial encoding is amissing key for implicit image function-based arbitrary-scale super-resolution[J].arXiv:2103.12716,2021.
[15]CHEN K,LI W,LEI S,et al.Continuous Remote Sensing Image Super-Resolution Based on Context Interaction in Implicit Function Space[J].IEEE Transactions on Geoscience and Remote Sensing,2023,61:1-15.
[16]WU H,NI N,ZHANG L.Learning Dynamic Scale Awarenessand Global Implicit Functions for Continuous-Scale Super-Resolution of Remote Sensing Images[J].IEEE Transactions on Geoscience and Remote Sensing,2023,61:1-16.
[17]DONG C,LOY C C,HE K,et al.Image super-resolution using deep convolutional networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(2):295-307.
[18]DONG C,LOY C C,TANG X.Accelerating the super-resolution convolutional neural network[C]//Computer Vision-ECCV 2016:14th European Conference,Amsterdam,The Netherlands,October 11-14,2016,Proceedings,Part II 14.Springer International Publishing,2016:391-407.
[19]KIM J,LEE J K,LEE K M.Accurate image super-resolutionusing very deep convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2016:1646-1654.
[20]LAI W S,HUANG J B,AHUJA N,et al.Deeplaplacian pyramid networks for fast and accurate super-resolution[C]//Procee-dings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:624-632.
[21]LEDIG C,THEIS L,HUSZÁR F,et al.Photo-realistic singleimage super-resolution using a generative adversarial network[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2017:4681-4690.
[22]WANG X,YU K,WU S,et al.Esrgan:Enhanced super-resolution generative adversarial networks[C]//Proceedings of the European Conference on Computer Vision(ECCV) Workshops.2018.
[23]GUO Y,CHEN J,WANG J,et al.Closed-loop matters:Dual regression networks for singleimage super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020:5407-5416.
[24]LIANG J,CAO J,SUN G,et al.SwinIR:Image restoration using Swin transformer[C]//International Conference on Computer Vision Workshops.2021:1833-1844.
[25]HU X,MU H,ZHANG X,et al.Meta SR:A magnification-arbitrary network for super-resolution[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).2019:1575-1584.
[26]LEE J,JIN K H.Local texture estimator for implicit representation function[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:1919-1928.
[27]LIU X,GAO S,ZENG B,et al.Implicit diffusion models for continuous super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023.
[28]DUAN M,QU L,LIU S,et al.Local Implicit Wavelet Transformer for Arbitrary-Scale Super-Resolution[C]//British Machine Vision Conference.2024.
[29]SITZMANN V,MARTEL J N P,BERGMAN A W,et al.Im-plicit neural representations with periodic activation functions[C]//Advances in Neural Information Processing Systems 33.2020:7462-7473.
[30]WU Z J,JIN Y H,YIK M.Neural fourier filter bank[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023.
[31]ATZMONM,LIPMAN Y.Sal:Sign agnostic learning of shapes from raw data[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2020.
[32]PARK J J,FLORENCE P,STRAUB J,et al.Deepsdf:Learning continuous signed distance functions for shape representation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2019:165-174.
[33]SHEN L Y,JOHN P,LEI X.NeRP:implicit neural representation learning with prior embedding for sparsely sampled image reconstruction[J].IEEE Transactions on Neural Networks and Learning Systems,2022,35(1):770-782.
[34]ZHANG Y,LI K,LI K,et al.Image super-resolution using very deep residual channel attention networks[C]//Proceedings of the European Conference on Computer Vision(ECCV).2018:286-301.
[35]DAI T,CAI J,ZHANG Y,et al.Second-order attention network for single image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2019:11065-11074.
[36]LIU B,SUN J,ZHU B,et al.MadFormer:multi-attention-driven image super-resolution method based on Transformer[J].Multimedia Systems,2024,30(2):78.
[37]DU N,ZHAO J.HyperFLFN:An Image Super-Resolution Mo-del Based on Attention Mechanism and Meta-Learning[J].Engineering Letters,2025,33(1):132-139.
[38]ZHAO L,GAO J,DENG D,et al.SSIR:Spatial shuffle multi-head self-attention for single image super-resolution[J].Pattern Recognition,2024,148:110195.
[39]QIN F,YAN K,WANG C,et al.LKFormer:Large kernel transformer for infrared image super-resolution[J].Multimedia Tools and Applications,2024,83(28):72063-72077.
[40]AGUSTSSON E,TIMOFTE R.Ntire 2017 challenge on single image super-resolution:Dataset and study[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.2017:126-135.
[41]ZEYDE R,ELAD M,PROTTER M.On single image scale-upusing sparse-representations[C]//Proceedings of the 7th International Conference on Curves and Surfaces.2010:711-730.
[42]MARTIN D,FOWLKES C,TAL D,et al.A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics[C]//Proceedings Eighth IEEE International Conference on Computer Vision(ICCV 2001).2001:416-423.
[43]HUANG J B,SINGH A,AHUJA N.Single image super-resolution from transformed self-exemplars[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.2015:5197-5206.
[44]YANG J,SHEN S,YUE H,et al.Implicit Transformer Network for Screen Content Image Continuous Super-Resolution[J].Advances in Neural Information Processing,2021,34:13304-13315.
[45]CAO J Z,WANG Q,XIAN Y,et al.Ciaosr:Continuous implicit attention-in-attention network for arbitrary-scale image super-resolution[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2023.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发