A fully-blind and fast image quality predictor with convolutional neural networks

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Image quality assessment (IQA) is an inherent problem in the field of image processing. Recently, deep learning-based image quality assessment has attracted increased attention, owing to its high prediction accuracy. In this paper, we propose a fully-blind and fast image quality predictor (FFIQP) using convolutional neural networks including two strategies. First, we propose a distortion clustering strategy based on the distribution function of intermediate-layer results in the convolutional neural network (CNN) to make IQA fully blind. Second, by analyzing the relationship between image saliency information and CNN prediction error, we utilize a pre-saliency map to skip the non-salient patches for IQA acceleration. Experimental results verify that our method can achieve the high accuracy (0.978) with subjective quality scores, outperforming existing IQA methods. Moreover, the proposed method is highly computationally appealing, achieving flexible complexity performance by assigning different thresholds in the saliency map.

Original languageEnglish
Pages (from-to)1557-1566
Number of pages10
JournalIEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences
VolumeE101A
Issue number9
DOIs
Publication statusPublished - 2018 Sep 1

Fingerprint

Image Quality Assessment
Image Quality
Image quality
Predictors
Neural Networks
Neural networks
Saliency Map
Saliency
Prediction Error
Patch
Image Processing
High Accuracy
Distribution Function
Distribution functions
Clustering
Verify
Image processing
Prediction
Experimental Results

Keywords

  • Convolutional neural networks
  • Distortion recognition
  • No-reference image quality assessment
  • Saliency map

ASJC Scopus subject areas

  • Signal Processing
  • Computer Graphics and Computer-Aided Design
  • Electrical and Electronic Engineering
  • Applied Mathematics

Cite this

@article{98933ef4e2b04983a35e1704278056c2,
title = "A fully-blind and fast image quality predictor with convolutional neural networks",
abstract = "Image quality assessment (IQA) is an inherent problem in the field of image processing. Recently, deep learning-based image quality assessment has attracted increased attention, owing to its high prediction accuracy. In this paper, we propose a fully-blind and fast image quality predictor (FFIQP) using convolutional neural networks including two strategies. First, we propose a distortion clustering strategy based on the distribution function of intermediate-layer results in the convolutional neural network (CNN) to make IQA fully blind. Second, by analyzing the relationship between image saliency information and CNN prediction error, we utilize a pre-saliency map to skip the non-salient patches for IQA acceleration. Experimental results verify that our method can achieve the high accuracy (0.978) with subjective quality scores, outperforming existing IQA methods. Moreover, the proposed method is highly computationally appealing, achieving flexible complexity performance by assigning different thresholds in the saliency map.",
keywords = "Convolutional neural networks, Distortion recognition, No-reference image quality assessment, Saliency map",
author = "Zhengxue Cheng and Masaru Takeuchi and Kenji Kanai and Jiro Katto",
year = "2018",
month = "9",
day = "1",
doi = "10.1587/transfun.E101.A.1557",
language = "English",
volume = "E101A",
pages = "1557--1566",
journal = "IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences",
issn = "0916-8508",
publisher = "Maruzen Co., Ltd/Maruzen Kabushikikaisha",
number = "9",

}

TY - JOUR

T1 - A fully-blind and fast image quality predictor with convolutional neural networks

AU - Cheng, Zhengxue

AU - Takeuchi, Masaru

AU - Kanai, Kenji

AU - Katto, Jiro

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Image quality assessment (IQA) is an inherent problem in the field of image processing. Recently, deep learning-based image quality assessment has attracted increased attention, owing to its high prediction accuracy. In this paper, we propose a fully-blind and fast image quality predictor (FFIQP) using convolutional neural networks including two strategies. First, we propose a distortion clustering strategy based on the distribution function of intermediate-layer results in the convolutional neural network (CNN) to make IQA fully blind. Second, by analyzing the relationship between image saliency information and CNN prediction error, we utilize a pre-saliency map to skip the non-salient patches for IQA acceleration. Experimental results verify that our method can achieve the high accuracy (0.978) with subjective quality scores, outperforming existing IQA methods. Moreover, the proposed method is highly computationally appealing, achieving flexible complexity performance by assigning different thresholds in the saliency map.

AB - Image quality assessment (IQA) is an inherent problem in the field of image processing. Recently, deep learning-based image quality assessment has attracted increased attention, owing to its high prediction accuracy. In this paper, we propose a fully-blind and fast image quality predictor (FFIQP) using convolutional neural networks including two strategies. First, we propose a distortion clustering strategy based on the distribution function of intermediate-layer results in the convolutional neural network (CNN) to make IQA fully blind. Second, by analyzing the relationship between image saliency information and CNN prediction error, we utilize a pre-saliency map to skip the non-salient patches for IQA acceleration. Experimental results verify that our method can achieve the high accuracy (0.978) with subjective quality scores, outperforming existing IQA methods. Moreover, the proposed method is highly computationally appealing, achieving flexible complexity performance by assigning different thresholds in the saliency map.

KW - Convolutional neural networks

KW - Distortion recognition

KW - No-reference image quality assessment

KW - Saliency map

UR - http://www.scopus.com/inward/record.url?scp=85053852160&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85053852160&partnerID=8YFLogxK

U2 - 10.1587/transfun.E101.A.1557

DO - 10.1587/transfun.E101.A.1557

M3 - Article

AN - SCOPUS:85053852160

VL - E101A

SP - 1557

EP - 1566

JO - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences

JF - IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences

SN - 0916-8508

IS - 9

ER -