Learning Frequency-aware Dynamic Network for Efficient Super-Resolution

  title={Learning Frequency-aware Dynamic Network for Efficient Super-Resolution},
  author={Wenbin Xie and Dehua Song and Chang Xu and Chunjing Xu and Hui Zhang and Yunhe Wang},
  journal={2021 IEEE/CVF International Conference on Computer Vision (ICCV)},
Deep learning based methods, especially convolutional neural networks (CNNs) have been successfully applied in the field of single image super-resolution (SISR). To obtain better fidelity and visual quality, most of existing networks are of heavy design with massive computation. However, the computation resources of modern mobile devices are limited, which cannot easily support the expensive cost. To this end, this paper explores a novel frequency-aware dynamic network for dividing the input… 

Figures and Tables from this paper

CADyQ: Content-Aware Dynamic Quantization for Image Super-Resolution

A novel CADyQ framework for SR networks that allocates optimal bits to local regions and layers adaptively based on the local contents of an input image to achieve high average bit-reduction with less accuracy loss is proposed.

GhostSR: Learning Ghost Features for Efficient Image Super-Resolution

Extensive experiments demonstrate that both the non-compact and lightweight SISR models embedded with the proposed method can achieve a comparable performance to that of their baselines with a large reduction of parameters, FLOPs and GPU inference latency.

ARM: Any-Time Super-Resolution Method

An Edge-to-PSNR lookup table is constructed that maps the edge score of an image patch to the PSNR performance for each subnet, together with a set of computation costs for the subnets.

Self-Supervised Learning for Real-World Super-Resolution from Dual Zoomed Observations

This paper presents a novel self-supervised learning approach for real-world image SR from observations at dual camera zooms (SelfDZSR), and takes the telephoto image instead of an additional high-resolution image as the supervision information and selects a center patch from it as the reference to super-resolve the corresponding short-focus image patch.

Incremental Cross-view Mutual Distillation for Self-supervised Medical CT Synthesis

A novel medical slice synthesis to increase the inter-slice resolution by introducing the incremental cross-view mutual distillation strategy to accomplish this task in the self-supervised learning manner and demonstrating the effectiveness of the proposed approach.



Fast and Accurate Single Image Super-Resolution via Information Distillation Network

This work proposes a deep but compact convolutional network to directly reconstruct the high resolution image from the original low resolution image and demonstrates that the proposed method is superior to the state-of-the-art methods, especially in terms of time performance.

Lightweight Image Super-Resolution with Information Multi-distillation Network

An adaptive cropping strategy (ACS) is developed to super-resolve block-wise image patches using the same well-trained model and performs favorably against the state-of-the-art SR algorithms in terms of visual quality, memory footprint, and inference time.

Efficient Super Resolution Using Binarized Neural Network

This work explores an network-binarization approach for SR tasks without sacrificing much reconstruction accuracy, and shows that binarized SR networks achieve comparable qualitative and quantitative results as their real-weight counterparts.

Efficient Residual Dense Block Search for Image Super-Resolution

An efficient residual dense block search algorithm with multiple objectives to hunt for fast, lightweight and accurate networks for image super-resolution models achieves better performance than the state-of-the-art methods with limited number of parameters and FLOPs.

Image Super-Resolution Using Very Deep Residual Channel Attention Networks

This work proposes a residual in residual (RIR) structure to form very deep network, which consists of several residual groups with long skip connections, and proposes a channel attention mechanism to adaptively rescale channel-wise features by considering interdependencies among channels.

Accelerating the Super-Resolution Convolutional Neural Network

This paper aims at accelerating the current SRCNN, and proposes a compact hourglass-shape CNN structure for faster and better SR, and presents the parameter settings that can achieve real-time performance on a generic CPU while still maintaining good performance.

Second-Order Attention Network for Single Image Super-Resolution

Experimental results demonstrate the superiority of the SAN network over state-of-the-art SISR methods in terms of both quantitative metrics and visual quality.

Image Super-Resolution via Attention Based Back Projection Networks

An Attention based Back Projection Network (ABPN) for image super-resolution, inspired by recent studies on attention models, and a Spatial Attention Block (SAB) to learn the cross-correlation across features at different layers.

Binarized Neural Network for Single Image Super Resolution

This work investigates the binary neural network-based SISR problem and proposes a novel model binarization method using a bit-accumulation mechanism (BAM) to approximate the full-precision convolution with a value accumulation scheme, which can gradually refine the precision of quantization along the direction of model inference.

Learning a Deep Convolutional Network for Image Super-Resolution

This work proposes a deep learning method for single image super-resolution (SR) that directly learns an end-to-end mapping between the low/high-resolution images and shows that traditional sparse-coding-based SR methods can also be viewed as a deep convolutional network.