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.

Fast and Robust Image Upsampling by Local Adaptive Gradient Field Sharpening Transform

Fast Image Super-Resolution via Local Adaptive Gradient Field Sharpening Transform.

Fast Image Upsampling via Adaptive Gradient Sharpening and Transformed- Texture Constraint

Highly Robust Real-Time Image Enhancement

Optimizing the Quality of Fourier Single-Pixel Imaging Via Generative Adversarial Network

L0 image smoothing via iterating truncated L1 gradient regularization

Improved Restoration Algorithm for Weakly Blurred and Strongly Noisy Image

Image Copy-and-paste with Optimized Gradient

Content Adaptive Constraint Based Image Upsampling.

Adaptive Sampling and Reconstruction for Gradient-Domain Rendering

High Resolution Local Structure-Constrained Image Upsampling

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

Gradient-based adaptive interpolation in super-resolution image restoration

Detail-preserving Joint Image Upsampling

Robust High Quality Image Guided Depth Upsampling

Multilevel Modified Finite Radon Transform Network for Image Upsampling

Iterative Projection Reconstruction for Fast and Efficient Image Upsampling

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

Faster Gradient Descent and the Efficient Recovery of Images

A Refreshed Similarity-based Upsampler for Direct High-Ratio Feature Upsampling

Depth Super-Resolution Using Joint Adaptive Weighted Least Squares and Patching Gradient.