Video inpainting and implant via diversified temporal continuations

Timothy K. Shih, Nick C. Tang, Wei Sung Yeh, Ta Jen Chen, Wonjun Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

17 Citations (Scopus)

Abstract

Recent interesting issues in video inpainting are defect removal and object removal. We take one more step to replace the removed objects in a video sequence by implanting objects from another video. Before implant, we improve an exemplar-based image inpainting algorithm by using a new patch matching strategy which incorporates edge properties. The data term used in a priority computation of candidate patches is also redefined. We take varieties of temporal continuations of foreground and background into consideration. A motion compensated inpainting procedure is then proposed. The inpainted video backgrounds are visually pleasant with smooth transitions. A simple tracking algorithm is then used to produce a foreground video, which is implanted into the inpainted background video. Our results are available at http://www.mine.tku.edu.tw/inpainting.

Original languageEnglish
Title of host publicationProceedings of the 14th Annual ACM International Conference on Multimedia, MM 2006
Pages133-136
Number of pages4
DOIs
Publication statusPublished - 2006
Event14th Annual ACM International Conference on Multimedia, MM 2006 - Santa Barbara, CA, United States
Duration: 2006 Oct 232006 Oct 27

Publication series

NameProceedings of the 14th Annual ACM International Conference on Multimedia, MM 2006

Other

Other14th Annual ACM International Conference on Multimedia, MM 2006
Country/TerritoryUnited States
CitySanta Barbara, CA
Period06/10/2306/10/27

Keywords

  • Object removal
  • Object tracking
  • Temporal continuations
  • Video implant
  • Video inpainting

ASJC Scopus subject areas

  • General Computer Science
  • Media Technology

Fingerprint

Dive into the research topics of 'Video inpainting and implant via diversified temporal continuations'. Together they form a unique fingerprint.

Cite this