Decouple Learning for Parameterized Image Operators (2018-07-21T00:00:00.000000Z)

TL;DR

A new decouple learning algorithm to learn from the operator parameters to dynamically adjust the weights of a deep network for image operators, denoted as the base network is proposed.

Abstract

Many different deep networks have been used to approximate, accelerate or improve traditional image operators, such as image smoothing, super-resolution and denoising. Among these traditional operators, many contain parameters which need to be tweaked to obtain the satisfactory results, which we refer to as “parameterized image operators”. However, most existing deep networks trained for these operators are only designed for one specific parameter configuration, which does not meet the needs of real scenarios that usually require flexible parameters settings. To overcome this limitation, we propose a new decouple learning algorithm to learn from the operator parameters to dynamically adjust the weights of a deep network for image operators, denoted as the base network. The learned algorithm is formed as another network, namely the weight learning network, which can be end-to-end jointly trained with the base network. Experiments demonstrate that the proposed framework can be successfully applied to many traditional parameterized image operators. We provide more analysis to better understand the proposed framework, which may inspire more promising research in this direction. Our codes and models have been released in https://github.com/fqnchina/DecoupleLearning.

Authors

Lu Yuan

7 papers

Dongdong Chen

7 papers

G. Hua

6 papers

TL;DR

Abstract

Authors

References53 items

S3D: Single Shot multi-Span Detector via Fully 3D Convolutional Networks

Deep exemplar-based colorization

Generating Diverse and Accurate Visual Captions by Comparative Adversarial Learning

Revisiting RCNN: On Awakening the Classification Power of Faster RCNN

Stereoscopic Neural Style Transfer

Density-Aware Single Image De-raining Using a Multi-stream Dense Network

DA-GAN: Instance-Level Image Translation by Deep Attention Generative Adversarial Networks

xUnit: Learning a Spatial Activation Function for Efficient Image Restoration

Global Versus Localized Generative Adversarial Nets

Deep Scene Image Classification with the MFAFVNet

Robust emotion recognition from low quality and low bit rate video: A deep learning approach

Fast Image Processing with Fully-Convolutional Networks

A Generic Deep Architecture for Single Image Reflection Removal and Image Smoothing

Temporal Context Network for Activity Localization in Videos

MemNet: A Persistent Memory Network for Image Restoration

FASON: First and Second Order Information Fusion Network for Texture Recognition

Removing Rain from Single Images via a Deep Detail Network

Adversarial Ranking for Language Generation

StyleBank: An Explicit Representation for Neural Image Style Transfer

Coherent Online Video Style Transfer

Learned Optimizers that Scale and Generalize

Revisiting Deep Intrinsic Image Decompositions

Improved Texture Networks: Maximizing Quality and Diversity in Feed-Forward Stylization and Texture Synthesis

Learning to Learn for Global Optimization of Black Box Functions

Learning Recursive Filters for Low-Level Vision via a Hybrid Neural Network

HyperNetworks

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

Clearing the Skies: A Deep Network Architecture for Single-Image Rain Removal

Learning to learn by gradient descent by gradient descent

Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections

Multi-Scale Context Aggregation by Dilated Convolutions

Accurate Image Super-Resolution Using Very Deep Convolutional Networks

Deep Edge-Aware Filters

Compression Artifacts Reduction by a Deep Convolutional Network

Rolling Guidance Filter

Learning a Deep Convolutional Network for Image Super-Resolution

Weighted Nuclear Norm Minimization with Application to Image Denoising

100+ Times Faster Weighted Median Filter (WMF)

Structure-preserving image smoothing via region covariances

Structure extraction from texture via relative total variation

Image smoothing via L0 gradient minimization

From learning models of natural image patches to whole image restoration

Image Super-Resolution Via Sparse Representation

Edge-preserving decompositions for multi-scale tone and detail manipulation

Image super-resolution using gradient profile prior

Synthesis and evaluation of linear motion transitions

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

Image Denoising Via Sparse and Redundant Representations Over Learned Dictionaries

A non-local algorithm for image denoising

Learning to Control Fast-Weight Memories: An Alternative to Dynamic Recurrent Networks

Feature-oriented image enhancement using shock filters

Decouple Learning for Parameterized Image Operators

Bayesian Image Super-Resolution

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names