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.

A Practical And Adaptive Framework For Super-Resolution

Spatially Adaptive Block-Based Super-Resolution

Practical super-resolution from dynamic video sequences

Efficient Image Super-Resolution Integration

A Single-Image Super-Resolution Method Based on Progressive-Iterative Approximation

A multi-frame adaptive super-resolution method using double channel and regional pixel information

A Multi-Frame Image Super-Resolution Method.

Image Super-Resolution Framework with Multi-Channel Constraints.

A novel approach to multi-frame image super resolution using an innovative filter and pixel selection mechanism

Multi-frame Super Resolution Using Refined Exploration of Extensive Self-examples.

Analysis of Active Superresolution Imaging Techniques for Realtime Applications

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

Fast Convergence Strategy for Multi-Image Superresolution via Adaptive Line Search.

Image super-resolution via adaptive filtering and regularization

A Robust Adaptive Super-Resolution Method

Adaptive algorithm of super-resolution image reconstruction

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

A Feature Reuse Framework with Texture-adaptive Aggregation for Reference-based Super-Resolution

KernelFlexSR: a self-adaptive super-resolution algorithm with multi-path convolution and residual network for dynamic kernel enhancement

An Adaptive Super-Resolution Method Based on Regional Pixel Information and Ringing Artifacts Suppression

Task Decoupled Framework for Reference-based Super-Resolution