Online crowdsourcing subjective image quality assessment

Qianqian Xu*, Qingming Huang, Yuan Yao

*Corresponding author for this work

Research output: Chapter in Book/Conference Proceeding/ReportConference Paper published in a bookpeer-review

61 Citations (Scopus)

Abstract

Recently, HodgeRank on random graphs has been proposed as an effective framework for multimedia quality assessment problem based on paired comparison method. With the random design on large graphs, it is particularly suitable for large scale crowdsourcing experiments on Internet. However, to make it more practical toward this purpose, it is necessary to develop online algorithms to deal with sequential or streaming data. In this paper, we propose an online rating scheme based on HodgeRank on random graphs, to assess image quality when assessors and image pairs enter the system in a sequential way in a crowdsourceable scenario. The scheme is shown in both theory and experiments to be effective by exhibiting similar performance to batch learning under the Erdös-Rényi random graph model for sampling. It enables us to derive global rating and monitor intrinsic inconsistency in the real time. We demonstrate the effectiveness of the proposed framework on LIVE and IVC databases.

Original languageEnglish
Title of host publicationMM 2012 - Proceedings of the 20th ACM International Conference on Multimedia
Pages359-368
Number of pages10
DOIs
Publication statusPublished - 2012
Externally publishedYes
Event20th ACM International Conference on Multimedia, MM 2012 - Nara, Japan
Duration: 29 Oct 20122 Nov 2012

Publication series

NameMM 2012 - Proceedings of the 20th ACM International Conference on Multimedia

Conference

Conference20th ACM International Conference on Multimedia, MM 2012
Country/TerritoryJapan
CityNara
Period29/10/122/11/12

Keywords

  • crowdsourcing
  • hodgerank
  • online
  • paired comparison
  • persistent homology
  • random graphs
  • subjective image quality assessment
  • topology evolution
  • triangular curl

Fingerprint

Dive into the research topics of 'Online crowdsourcing subjective image quality assessment'. Together they form a unique fingerprint.

Cite this