LAPAR: Linearly-Assembled Pixel-Adaptive Regression Network for Single Image Super-resolution and Beyond (2021-05-21T00:00:00.000000Z)

TL;DR

A linearly-assembled pixel-adaptive regression network (LAPAR) is proposed, which casts the direct LR to HR mapping learning into a linear coefficient regression task over a dictionary of multiple predefined filter bases, which renders the model highly lightweight and easy to optimize while achieving state-of-the-art results on SISR benchmarks.

Abstract

Single image super-resolution (SISR) deals with a fundamental problem of upsampling a low-resolution (LR) image to its high-resolution (HR) version. Last few years have witnessed impressive progress propelled by deep learning methods. However, one critical challenge faced by existing methods is to strike a sweet spot of deep model complexity and resulting SISR quality. This paper addresses this pain point by proposing a linearly-assembled pixel-adaptive regression network (LAPAR), which casts the direct LR to HR mapping learning into a linear coefficient regression task over a dictionary of multiple predefined filter bases. Such a parametric representation renders our model highly lightweight and easy to optimize while achieving state-of-the-art results on SISR benchmarks. Moreover, based on the same idea, LAPAR is extended to tackle other restoration tasks, e.g., image denoising and JPEG image deblocking, and again, yields strong performance. The code is available at https://github.com/dvlab-research/Simple-SR.

Authors

Jiangbo Lu

3 papers

Jiaya Jia

23 papers

Lu Qi

5 papers

TL;DR

Abstract

Authors

References57 items

Unfolding the Alternating Optimization for Blind Super Resolution

Deep Unfolding Network for Image Super-Resolution

Second-Order Attention Network for Single Image Super-Resolution

Lightweight Image Super-Resolution with Adaptive Weighted Learning Network

Feedback Network for Image Super-Resolution

Meta-SR: A Magnification-Arbitrary Network for Super-Resolution

Fast, Accurate and Lightweight Super-Resolution with Neural Architecture Search

ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks

Image Super-Resolution Using Very Deep Residual Channel Attention Networks

Deep Video Super-Resolution Network Using Dynamic Upsampling Filters Without Explicit Motion Compensation

A Fully Progressive Approach to Single-Image Super-Resolution

Fast, Accurate, and, Lightweight Super-Resolution with Cascading Residual Network

Deep Back-Projection Networks for Super-Resolution

Learning a Single Convolutional Super-Resolution Network for Multiple Degradations

Burst Denoising with Kernel Prediction Networks

Spatially-Adaptive Filter Units for Deep Neural Networks

BLADE: Filter learning for general purpose computational photography

Image Super-Resolution Using Dense Skip Connections

MemNet: A Persistent Memory Network for Image Restoration

NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study

A Deep Convolutional Neural Network with Selection Units for Super-Resolution

Kernel-predicting convolutional networks for denoising Monte Carlo renderings

Enhanced Deep Residual Networks for Single Image Super-Resolution

Image Super-Resolution via Deep Recursive Residual Network

Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution

Video Frame Interpolation via Adaptive Convolution

Video Frame Synthesis Using Deep Voxel Flow

Towards deep compositional networks

Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

Accelerating the Super-Resolution Convolutional Neural Network

Visual Dynamics: Probabilistic Future Frame Synthesis via Cross Convolutional Networks

Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network

RAISR: Rapid and Accurate Image Super Resolution

Dynamic Filter Networks

Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks

Deeply-Recursive Convolutional Network for Image Super-Resolution

Accurate Image Super-Resolution Using Very Deep Convolutional Networks

Sketch-based manga retrieval using manga109 dataset

Single image super-resolution from transformed self-exemplars

Learning a Deep Convolutional Network for Image Super-Resolution

Anchored Neighborhood Regression for Fast Example-Based Super-Resolution

Low-Complexity Single-Image Super-Resolution based on Nonnegative Neighbor Embedding

Coupled Dictionary Training for Image Super-Resolution

From learning models of natural image patches to whole image restoration

Image Super-Resolution Via Sparse Representation

On Single Image Scale-Up Using Sparse-Representations

An introduction to Total Variation for Image Analysis

Coherent line drawing

Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering

Real-time video abstraction

Image quality assessment: from error visibility to structural similarity

A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics

Total variation based image restoration with free local constraints

International Journal of Computer Vision manuscript No. (will be inserted by the editor) Dictionary Learning for Deblurring and Digital Zoom

Image Abstraction by Structure Adaptive Filtering

Human facial illustrations: Creation and psychophysical evaluation

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 12. NO. 7. JULY 1990 Scale-Space and Edge Detection Using Anisotropic Diffusion

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names