Publication Details

Learning to Predict Localized Distortions in Rendered Images

ČADÍK, M.; HERZOG, R.; MANTIUK, R.; MANTIUK, R.; MYSZKOWSKI, K.; SEIDEL, H. Learning to Predict Localized Distortions in Rendered Images. COMPUTER GRAPHICS FORUM, 2013, vol. 2013, no. 7, p. 401-410. ISSN: 0167-7055.
Czech title
Predikce chyb v syntetických obrazech pomocí strojového učení
Type
journal article
Language
English
Authors
Čadík Martin, doc. Ing., Ph.D. (DCGM)
Herzog Robert
Mantiuk Rafał
Mantiuk Radosław
Myszkowski Karol
Seidel Hans-Peter
URL
Keywords

image quality assessment, feature vectors, machine learning, rendering

Abstract

In this work, we present an analysis of feature descriptors for objective image quality assessment. We explore a large space of possible features including components of existing image quality metrics as well as many traditional computer vision and statistical features. Additionally, we propose new features motivated by human perception and we analyze visual saliency maps acquired using an eye tracker in our user experiments. The discriminative power of the features is assessed by means of a machine learning framework revealing the importance of each feature for image quality assessment task. Furthermore, we propose a new data-driven full-reference image quality metric which outperforms current state-of-the-art metrics. The metric was trained on subjective ground truth data combining two publicly available datasets. For the sake of completeness we create a new testing synthetic dataset including experimentally measured subjective distortion maps. Finally, using the same machine-learning framework we optimize the parameters of popular existing metrics.

Published
2013
Pages
401–410
Journal
COMPUTER GRAPHICS FORUM, vol. 2013, no. 7, ISSN 0167-7055
DOI
UT WoS
000327310800042
BibTeX
@article{BUT103580,
  author="Martin {Čadík} and Robert {Herzog} and Rafał {Mantiuk} and Radosław {Mantiuk} and Karol {Myszkowski} and Hans-Peter {Seidel}",
  title="Learning to Predict Localized Distortions in Rendered Images",
  journal="COMPUTER GRAPHICS FORUM",
  year="2013",
  volume="2013",
  number="7",
  pages="401--410",
  doi="10.1111/cgf.12248",
  issn="0167-7055",
  url="http://cadik.posvete.cz"
}
Back to top