Skip to main navigation Skip to search Skip to main content

Online Video Object Segmentation Based on Region and Edge Consistency

  • Jingjing Ma
  • , Qingjie Zhao
  • , Peng Lv
  • , Jimmy T. Mbelwa
  • , Hao Liu
  • , Jianwei Zhang
  • Beijing Institute of Technology
  • University of Hamburg

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper proposes a novel online video object segmentation method by using region and edge consistency information. Motivated by the fact that the object edge information is similar between adjacent frames, we firstly generate object proposal based on region and edge consistency, which strengthens the confidence of regions nearby the obvious edges. Subsequently, we fuse optical flow into Markov Random Field (MRF) model to accurately segment moving object. Given an annotation frame for the initial object region, we can process video frames online within a short time. The experiments on benchmark video dataset demonstrate that the proposed method outperforms the existing state-of-the-art methods, especially in the scenarios with similar objects.

Original languageEnglish
Title of host publicationProceedings - 2018 IEEE International Conference on Big Data and Smart Computing, BigComp 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages333-339
Number of pages7
ISBN (Electronic)9781538636497
DOIs
Publication statusPublished - 25 May 2018
Event2018 IEEE International Conference on Big Data and Smart Computing, BigComp 2018 - Shanghai, China
Duration: 15 Jan 201818 Jan 2018

Publication series

NameProceedings - 2018 IEEE International Conference on Big Data and Smart Computing, BigComp 2018

Conference

Conference2018 IEEE International Conference on Big Data and Smart Computing, BigComp 2018
Country/TerritoryChina
CityShanghai
Period15/01/1818/01/18

Keywords

  • MRF
  • edge consistency.
  • object segmentation
  • superpixel
  • supervoxelm

Fingerprint

Dive into the research topics of 'Online Video Object Segmentation Based on Region and Edge Consistency'. Together they form a unique fingerprint.

Cite this