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.

Second-Order Directional Total Generalized Variation Regularization for Image Super-resolution

Single-image super-resolution using directional total variation regularization and alternating direction method of multiplier solver

Single Image Super-Resolution Using Combined Total Variation Regularization by Split Bregman Iteration

Single Image Superresolution Via Directional Group Sparsity and Directional Features.

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

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

A Multi-Frame Image Super-Resolution Method.

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

Super-resolution Reconstruction Based on Geometric Dictionary Learning and Coupled Regularization

A Regularized Super Resolution Algorithm for Generalized Gaussian Noise

GDSSR: Toward Real-World Ultra-High-Resolution Image Super-Resolution

Single Image Super-Resolution Via Locally Regularized Anchored Neighborhood Regression and Nonlocal Means.

Regularized Super-resolution Reconstruction Based on Bilateral Total Variation Model with Adaptive Constraints

Manifold regularized sparse support regression for single image super-resolution

SRWGANTV: Image Super-Resolution Through Wasserstein Generative Adversarial Networks with Total Variational Regularization

Image Super-Resolutton Reconstruction Based On Adaptive Interpolation Norm Regularization

Single Image Super-Resolution Based on Nonlocal Sparse and Low-Rank Regularization.

Unified Regularization Framework for Blind Image Super-Resolution

Image Super-Resolution Using Multi-Layer Support Vector Regression

Generalized Joint Kernel Regression and Adaptive Dictionary Learning for Single-Image Super-Resolution.

Orthogonally Regularized Deep Networks For Image Super-resolution