Abstract:In this paper, the authors propose a joint SR model (JIRSR) that can effectively handle multiple degradations in a single model. Specifically, the authors build the parallel transformer and convolutional neural network branches that complement each other through bidirectional feature fusion. Extensive experiments on classical super‐resolution, denoising, and multiple degradation removal demonstrate that the authors' JIRSR achieves state‐of‐the‐art performance on public benchmarks. While recent years have witnessed the unprecedented success of deep convolutional neural networks (CNNs) and vision transformers in single‐image super‐resolution (SISR), the degradation assumptions are simple and usually bicubic downsampling. Thus, their performances will drop dramatically when the actual degradation does not match this assumption, and they lack the capability to handle multiple degradations (e.g. Gaussian noise, bicubic downsizing, and salt & pepper noise). To address the issues, in this paper, the authors propose a joint SR model (JIRSR) that can effectively handle multiple degradations in a single model. Specifically, the authors build the parallel Transformer and CNN branches that complement each other through bidirectional feature fusion. Moreover, the authors also adopt a random permutation of different kinds of noise and resizing operations to build the training datasets. Extensive experiments on classical SR, denoising, and multiple degradation removal demonstrate that the authors' JIRSR achieves state‐of‐the‐art (SOTA) performance on public benchmarks. Concretely, the authors' JIRSR outperforms the second‐best model by 0.23 to 0.74 dB for multiple degradations removal and is 0.20 to 0.36 dB higher than the SOTA methods on the Urban100 dataset under the ×4 SR task.

PRED: A PARALLEL NETWORK FOR HANDLING MULTIPLE DEGRADATIONS VIA SINGLE MODEL IN SINGLE IMAGE SUPER-RESOLUTION

Epistemic-Uncertainty-Based Divide-and-Conquer Network for Single-Image Super-Resolution

Conditional Hyper-Network for Blind Super-Resolution with Multiple Degradations

Orientation-Aware Deep Neural Network for Real Image Super-Resolution.

DeeDSR: Towards Real-World Image Super-Resolution via Degradation-Aware Stable Diffusion

Fast and Accurate Single Image Super-Resolution Via an Energy-Aware Improved Deep Residual Network.

A joint image super‐resolution network for multiple degradations removal via complementary transformer and convolutional neural network

E2-RealSR: efficient and effective real-world super-resolution network based on partial degradation modulation

Efficient blind super-resolution imaging via adaptive degradation-aware estimation

Single Image Super Resolution based on a Modified U-net with Mixed Gradient Loss

Non-local degradation modeling for spatially adaptive single image super-resolution

Unsupervised Degradation Learning for Single Image Super-Resolution

A Fast And Accurate Super-Resolution Network Using Progressive Residual Learning

VDUP1: a potential mediator of expansion-induced lung growth and epithelial cell differentiation in the ovine fetus.

DDNSR: a dual-input degradation network for real-world super-resolution

Super-resolution Reconstruction of Single Image for Latent features

A method of degradation mechanism-based unsupervised remote sensing image super-resolution

Claudin‐derived peptides are internalized via specific endocytosis pathways

Deep Networks for Image and Video Super-Resolution

Gradient Enhanced Dual Regression Network: Perception-Preserving Super-Resolution for Multi-Sensor Remote Sensing Imagery

RestoreDet: Degradation Equivariant Representation for Object Detection in Low Resolution Images