Abstract:Real-SR endeavors to produce high-resolution images with rich details while mitigating the impact of multiple degradation factors. Although existing methods have achieved impressive achievements in detail recovery, they still fall short when addressing regions with complex gradient arrangements due to the intensity-based linear weighting feature extraction manner. Moreover, the stochastic artifacts introduced by degradation cues during the imaging process in real LR increase the disorder of the overall image details, further complicating the perception of intrinsic gradient arrangement. To address these challenges, we innovatively introduce kernel-wise differential operations within the convolutional kernel and develop several learnable directional gradient convolutions. These convolutions are integrated in parallel with a novel linear weighting mechanism to form an Adaptive Directional Gradient Convolution (DGConv), which adaptively weights and fuses the basic directional gradients to improve the gradient arrangement perception capability for both regular and irregular textures. Coupled with DGConv, we further devise a novel equivalent parameter fusion method for DGConv that maintains its rich representational capabilities while keeping computational costs consistent with a single Vanilla Convolution (VConv), enabling DGConv to improve the performance of existing super-resolution networks without incurring additional computational expenses. To better leverage the superiority of DGConv, we further develop an Adaptive Information Interaction Block (AIIBlock) to adeptly balance the enhancement of texture and contrast while meticulously investigating the interdependencies, culminating in the creation of a DGPNet for Real-SR through simple stacking. Comparative results with 15 SOTA methods across three public datasets underscore the effectiveness and efficiency of our proposed approach.

The Improved Deep Plug-and-play Super-Resolution with Residual-in-residual Dense Block for Arbitrary Blur Kernels

Deep Plug-and-Play Super-Resolution for Arbitrary Blur Kernels

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

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

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

Two-Stage Deep Single Image Super-Resolution with Multiple Blur Kernels for the Internet of Things

Blind Super-Resolution on Remote Sensing Images with Blur Kernel Prediction.

A New Dataset and Framework for Real-World Blurred Images Super-Resolution

A New Super Resolution Framework Based on Multi-Task Learning for Remote Sensing Images

Robust Dual Images Super-Resolution.

Meta-learning based blind image super-resolution approach to different degradations

Revisiting Single Image Super-Resolution Under Internet Environment: Blur Kernels and Reconstruction Algorithms

Ultra Sharp : Study of Single Image Super Resolution using Residual Dense Network

Dense Contrastive Learning and Depth Dynamic Aggregation for Reference-based Super-Resolution.

Image super-resolution via dynamic network

Dual-Path Large Kernel Learning and Its Applications in Single-Image Super-Resolution

LAPAR: Linearly-Assembled Pixel-Adaptive Regression Network for Single Image Super-Resolution and Beyond

Efficient Real-world Image Super-Resolution Via Adaptive Directional Gradient Convolution

Iterative Network for Image Super-Resolution

DDet: Dual-path Dynamic Enhancement Network for Real-World Image Super-Resolution

Dynamic feature distillation and pyramid split large kernel attention network for lightweight image super-resolution