Lightweight Single-Image Super-Resolution Network with Attentive Auxiliary Feature Learning

Wang, Xuehui; Wang, Qing; Zhao, Yuzhi; Yan, Junchi; Fan, Lei; Chen, Long

doi:10.1007/978-3-030-69532-3_17

Cited by 24 publications

(24 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The BI degradation model has been widely used to obtain LR images in the image SR tasks. In order to demonstrate the effectiveness of the RTAN, we compared it with 16 state-of-the-art CNN-based SR methods, including SRMDNF [7], NLRN [32], EDSR [17], DBPN [73], NDRCN [35], ACNet [38], FALSR-A [37], OISR-RK2-s [34], MCAN [47], A 2 F-SD [48], A2N-M [63], DeFiAN S [61], IMDN [33], SMSR [36], PAN [59], MGAN [62], RNAN [55].…”

Section: Results With Bicubic (Bi) Degradation Modelmentioning

confidence: 99%

“…In order to demonstrate the powerful reconstruction ability of the proposed method with BD degradation model, we compare the RTAN with 14 state-of-the-art CNN-based models, i.e., SPMSR [4], SRCNN [5], FSRCNN [74], VDSR [12], SRMD [7], EDSR [17], RDN [16], IRCNN [75], SRFBN [76], RCAN [6], A 2 F-SD [48], IMDN [33], DeFiAN S [61], PAN [59], and MGAN [62].…”

Section: Results With Blur-downscale (Bd) Degradation Modelmentioning

confidence: 99%

“…Recently, some researchers have proposed attention-based models to improve the SR performance. Several works, such as RCAN [6], SAN [21], MCAN [47], A 2 F [48], LatticeNet [49], and DIN [50], introduce the channel attention (CA) mechanism to SR which makes the network learn more useful features. To learn more discriminative features, some researchers utilize both the channel attention and spatial attention, such as HRAN [51], MIRNet [52], CSFM [53], and BAM [54].…”

Section: Attention Mechanismmentioning

confidence: 99%

See 2 more Smart Citations

Residual Triplet Attention Network for Single-Image Super-Resolution

Huang

Wang

et al. 2021

Electronics

View full text Add to dashboard Cite

Single-image super-resolution (SISR) techniques have been developed rapidly with the remarkable progress of convolutional neural networks (CNNs). The previous CNNs-based SISR techniques mainly focus on the network design while ignoring the interactions and interdependencies between different dimensions of the features in the middle layers, consequently hindering the powerful learning ability of CNNs. In order to address this problem effectively, a residual triplet attention network (RTAN) for efficient interactions of the feature information is proposed. Specifically, we develop an innovative multiple-nested residual group (MNRG) structure to improve the learning ability for extracting the high-frequency information and train a deeper and more stable network. Furthermore, we present a novel lightweight residual triplet attention module (RTAM) to obtain the cross-dimensional attention weights of the features. The RTAM combines two cross-dimensional interaction blocks (CDIBs) and one spatial attention block (SAB) base on the residual module. Therefore, the RTAM is not only capable of capturing the cross-dimensional interactions and interdependencies of the features, but also utilizing the spatial information of the features. The simulation results and analysis show the superiority of the proposed RTAN over the state-of-the-art SISR networks in terms of both evaluation metrics and visual results.

show abstract

Section: Results With Bicubic (Bi) Degradation Modelmentioning

confidence: 99%

Section: Results With Blur-downscale (Bd) Degradation Modelmentioning

confidence: 99%

Section: Attention Mechanismmentioning

confidence: 99%

See 1 more Smart Citation

Residual Triplet Attention Network for Single-Image Super-Resolution

Huang

Wang

et al. 2021

Electronics

View full text Add to dashboard Cite

show abstract

“…Besides that, Wang et al [32] proposed local and global feature-fusion modules within and across residual blocks, respectively, which is efficient and can effectively improve the performance. Instead of directly fusing features extracted from intermediate layers through summation, Wang et al [33] proposed to project the extracted features into a shared space, and then fuse them in this common space for the reconstruction. The information distillation block proposed by IDN [14] is another efficient compact module for SISR, which splits the features generated into two parts in a residual block.…”

Section: Related Work 21 Deep Lightweight Sisr Modelsmentioning

confidence: 99%

“…Wang et al [32] proposed a local feature-fusion strategy within the residual block and a global feature-fusion strategy across the residual blocks to improve network efficiency. Wang et al [33] proposed to project the generated features from the intermediate layers into a shared space for feature fusion, which can further improve the performance. The information distillation block proposed in [13,14] is another efficient and compact module, which can progressively extract features for the reconstruction, leading to a better trade-off between performance and model parameters.…”

Section: Introductionmentioning

confidence: 99%

Self-feature Learning

Xiao

Zhao

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Deep learning-based models have achieved unprecedented performance in single image super-resolution (SISR). However, existing deep learning-based models usually require high computational complexity to generate high-quality images, which limits their applications in edge devices, e.g., mobile phones. To address this issue, we propose a dynamic, channel-agnostic filtering method in this paper. The proposed method not only adaptively generates convolutional kernels based on the local information of each position, but also can significantly reduce the cost of computing the interchannel redundancy. Based on this, we further propose a simple, yet effective, deep lightweight model for SISR. Experiment results show that our proposed model outperforms other state-of-the-art deep lightweight SISR models, leading to the best trade-off between the performance and the number of model parameters. CCS CONCEPTS• Computing methodologies → Image processing; Computational photography.

show abstract