Neural network-based full-reference image quality assessment

Sebastian Bosse; Dominique Maniry; Klaus Muller; Thomas Wiegand; Wojciech Samek

doi:10.1109/PCS.2016.7906376

Neural network-based full-reference image quality assessment

Sebastian Bosse, Dominique Maniry, Klaus Muller, Thomas Wiegand, Wojciech Samek

Department of Brain and Cognitive Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

14 Citations (Scopus)

Abstract

This paper presents a full-reference (FR) image quality assessment (IQA) method based on a deep convolutional neural network (CNN). The CNN extracts features from distorted and reference image patches and estimates the perceived quality of the distorted ones by combining and regressing the feature vectors using two fully connected layers. The CNN consists of 12 convolution and max-pooling layers; activation is done by a rectifier activation function (ReLU). The overall IQA score is computed by aggregating the patch quality estimates. Three different feature combination methods and two aggregation approaches are proposed and evaluated in this paper. Experiments are performed on the LIVE and TID2013 databases. On both databases linear Pearson correlations superior to state-of-the-art IQA methods are achieved.

Original language	English
Title of host publication	2016 Picture Coding Symposium, PCS 2016
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781509059669
DOIs	https://doi.org/10.1109/PCS.2016.7906376
Publication status	Published - 2017 Apr 19
Event	2016 Picture Coding Symposium, PCS 2016 - Nuremberg, Germany Duration: 2016 Dec 4 → 2016 Dec 7

Other

Other	2016 Picture Coding Symposium, PCS 2016
Country/Territory	Germany
City	Nuremberg
Period	16/12/4 → 16/12/7

ASJC Scopus subject areas

Media Technology
Signal Processing

Access to Document

10.1109/PCS.2016.7906376

Cite this

@inproceedings{719acf707b474aa8979cb7cc25bacc61,

title = "Neural network-based full-reference image quality assessment",

abstract = "This paper presents a full-reference (FR) image quality assessment (IQA) method based on a deep convolutional neural network (CNN). The CNN extracts features from distorted and reference image patches and estimates the perceived quality of the distorted ones by combining and regressing the feature vectors using two fully connected layers. The CNN consists of 12 convolution and max-pooling layers; activation is done by a rectifier activation function (ReLU). The overall IQA score is computed by aggregating the patch quality estimates. Three different feature combination methods and two aggregation approaches are proposed and evaluated in this paper. Experiments are performed on the LIVE and TID2013 databases. On both databases linear Pearson correlations superior to state-of-the-art IQA methods are achieved.",

author = "Sebastian Bosse and Dominique Maniry and Klaus Muller and Thomas Wiegand and Wojciech Samek",

year = "2017",

month = apr,

day = "19",

doi = "10.1109/PCS.2016.7906376",

language = "English",

booktitle = "2016 Picture Coding Symposium, PCS 2016",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

note = "2016 Picture Coding Symposium, PCS 2016 ; Conference date: 04-12-2016 Through 07-12-2016",

}

TY - GEN

T1 - Neural network-based full-reference image quality assessment

AU - Bosse, Sebastian

AU - Maniry, Dominique

AU - Muller, Klaus

AU - Wiegand, Thomas

AU - Samek, Wojciech

PY - 2017/4/19

Y1 - 2017/4/19

N2 - This paper presents a full-reference (FR) image quality assessment (IQA) method based on a deep convolutional neural network (CNN). The CNN extracts features from distorted and reference image patches and estimates the perceived quality of the distorted ones by combining and regressing the feature vectors using two fully connected layers. The CNN consists of 12 convolution and max-pooling layers; activation is done by a rectifier activation function (ReLU). The overall IQA score is computed by aggregating the patch quality estimates. Three different feature combination methods and two aggregation approaches are proposed and evaluated in this paper. Experiments are performed on the LIVE and TID2013 databases. On both databases linear Pearson correlations superior to state-of-the-art IQA methods are achieved.

AB - This paper presents a full-reference (FR) image quality assessment (IQA) method based on a deep convolutional neural network (CNN). The CNN extracts features from distorted and reference image patches and estimates the perceived quality of the distorted ones by combining and regressing the feature vectors using two fully connected layers. The CNN consists of 12 convolution and max-pooling layers; activation is done by a rectifier activation function (ReLU). The overall IQA score is computed by aggregating the patch quality estimates. Three different feature combination methods and two aggregation approaches are proposed and evaluated in this paper. Experiments are performed on the LIVE and TID2013 databases. On both databases linear Pearson correlations superior to state-of-the-art IQA methods are achieved.

UR - http://www.scopus.com/inward/record.url?scp=85019408658&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85019408658&partnerID=8YFLogxK

U2 - 10.1109/PCS.2016.7906376

DO - 10.1109/PCS.2016.7906376

M3 - Conference contribution

AN - SCOPUS:85019408658

BT - 2016 Picture Coding Symposium, PCS 2016

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2016 Picture Coding Symposium, PCS 2016

Y2 - 4 December 2016 through 7 December 2016

ER -

Neural network-based full-reference image quality assessment

Abstract

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this