Learning recursive filters for low-level vision via a hybrid neural network

Sifei Liu, Jinshan Pan, Ming Hsuan Yang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

33 Citations (Scopus)

Abstract

In this paper, we consider numerous low-level vision problems (e.g., edge-preserving filtering and denoising) as recursive image filtering via a hybrid neural network. The network contains several spatially variant recurrent neural networks (RNN) as equivalents of a group of distinct recursive filters for each pixel, and a deep convolutional neural network (CNN) that learns the weights of RNNs. The deep CNN can learn regulations of recurrent propagation for various tasks and effectively guides recurrent propagation over an entire image. The proposed model does not need a large number of convolutional channels nor big kernels to learn features for low-level vision filters. It is significantly smaller and faster in comparison with a deep CNN based image filter. Experimental results show that many low-level vision tasks can be effectively learned and carried out in real-time by the proposed algorithm.

Original languageEnglish
Title of host publicationComputer Vision - 14th European Conference, ECCV 2016, Proceedings
EditorsNicu Sebe, Bastian Leibe, Max Welling, Jiri Matas
PublisherSpringer Verlag
Pages560-576
Number of pages17
ISBN (Print)9783319464923
DOIs
Publication statusPublished - 2016 Jan 1

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9908 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Fingerprint

Neural Networks
Filter
Neural networks
Propagation
Image Filtering
Recurrent neural networks
Recurrent Neural Networks
Denoising
Filtering
Pixel
Pixels
Entire
kernel
Distinct
Real-time
Vision
Learning
Experimental Results
Model

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Liu, S., Pan, J., & Yang, M. H. (2016). Learning recursive filters for low-level vision via a hybrid neural network. In N. Sebe, B. Leibe, M. Welling, & J. Matas (Eds.), Computer Vision - 14th European Conference, ECCV 2016, Proceedings (pp. 560-576). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9908 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46493-0_34
Liu, Sifei ; Pan, Jinshan ; Yang, Ming Hsuan. / Learning recursive filters for low-level vision via a hybrid neural network. Computer Vision - 14th European Conference, ECCV 2016, Proceedings. editor / Nicu Sebe ; Bastian Leibe ; Max Welling ; Jiri Matas. Springer Verlag, 2016. pp. 560-576 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{bb902127a6804dcbbc405e1906b39c7b,
title = "Learning recursive filters for low-level vision via a hybrid neural network",
abstract = "In this paper, we consider numerous low-level vision problems (e.g., edge-preserving filtering and denoising) as recursive image filtering via a hybrid neural network. The network contains several spatially variant recurrent neural networks (RNN) as equivalents of a group of distinct recursive filters for each pixel, and a deep convolutional neural network (CNN) that learns the weights of RNNs. The deep CNN can learn regulations of recurrent propagation for various tasks and effectively guides recurrent propagation over an entire image. The proposed model does not need a large number of convolutional channels nor big kernels to learn features for low-level vision filters. It is significantly smaller and faster in comparison with a deep CNN based image filter. Experimental results show that many low-level vision tasks can be effectively learned and carried out in real-time by the proposed algorithm.",
author = "Sifei Liu and Jinshan Pan and Yang, {Ming Hsuan}",
year = "2016",
month = "1",
day = "1",
doi = "10.1007/978-3-319-46493-0_34",
language = "English",
isbn = "9783319464923",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "560--576",
editor = "Nicu Sebe and Bastian Leibe and Max Welling and Jiri Matas",
booktitle = "Computer Vision - 14th European Conference, ECCV 2016, Proceedings",
address = "Germany",

}

Liu, S, Pan, J & Yang, MH 2016, Learning recursive filters for low-level vision via a hybrid neural network. in N Sebe, B Leibe, M Welling & J Matas (eds), Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9908 LNCS, Springer Verlag, pp. 560-576. https://doi.org/10.1007/978-3-319-46493-0_34

Learning recursive filters for low-level vision via a hybrid neural network. / Liu, Sifei; Pan, Jinshan; Yang, Ming Hsuan.

Computer Vision - 14th European Conference, ECCV 2016, Proceedings. ed. / Nicu Sebe; Bastian Leibe; Max Welling; Jiri Matas. Springer Verlag, 2016. p. 560-576 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9908 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Learning recursive filters for low-level vision via a hybrid neural network

AU - Liu, Sifei

AU - Pan, Jinshan

AU - Yang, Ming Hsuan

PY - 2016/1/1

Y1 - 2016/1/1

N2 - In this paper, we consider numerous low-level vision problems (e.g., edge-preserving filtering and denoising) as recursive image filtering via a hybrid neural network. The network contains several spatially variant recurrent neural networks (RNN) as equivalents of a group of distinct recursive filters for each pixel, and a deep convolutional neural network (CNN) that learns the weights of RNNs. The deep CNN can learn regulations of recurrent propagation for various tasks and effectively guides recurrent propagation over an entire image. The proposed model does not need a large number of convolutional channels nor big kernels to learn features for low-level vision filters. It is significantly smaller and faster in comparison with a deep CNN based image filter. Experimental results show that many low-level vision tasks can be effectively learned and carried out in real-time by the proposed algorithm.

AB - In this paper, we consider numerous low-level vision problems (e.g., edge-preserving filtering and denoising) as recursive image filtering via a hybrid neural network. The network contains several spatially variant recurrent neural networks (RNN) as equivalents of a group of distinct recursive filters for each pixel, and a deep convolutional neural network (CNN) that learns the weights of RNNs. The deep CNN can learn regulations of recurrent propagation for various tasks and effectively guides recurrent propagation over an entire image. The proposed model does not need a large number of convolutional channels nor big kernels to learn features for low-level vision filters. It is significantly smaller and faster in comparison with a deep CNN based image filter. Experimental results show that many low-level vision tasks can be effectively learned and carried out in real-time by the proposed algorithm.

UR - http://www.scopus.com/inward/record.url?scp=84990026484&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990026484&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46493-0_34

DO - 10.1007/978-3-319-46493-0_34

M3 - Conference contribution

AN - SCOPUS:84990026484

SN - 9783319464923

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 560

EP - 576

BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings

A2 - Sebe, Nicu

A2 - Leibe, Bastian

A2 - Welling, Max

A2 - Matas, Jiri

PB - Springer Verlag

ER -

Liu S, Pan J, Yang MH. Learning recursive filters for low-level vision via a hybrid neural network. In Sebe N, Leibe B, Welling M, Matas J, editors, Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Springer Verlag. 2016. p. 560-576. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46493-0_34