Fast and robust varying affine stitching by proper feature point selection

Ling Rong, Yuli Sun, Ci Wang

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

Abstract

Image stitching algorithms are often proposed to connect the planar scenes from parallel free camera at the same physical location through parametric transforms. Occlusion inevitably exists and makes the parametric transform abruptly changed, if the cameras are installed on different locations. To solve this problem, smoothly varying affine stitching (SVAS) field has been proposed, but its performance and execution speed may be worse along with the increasing number of feature points. Based on the stability of its cost function, two rules are designed in this paper to do feature point selection, with consideration of their even distribution and similarity. Under the designed criterions, stitching performance is significantly improved.

Original languageEnglish
Title of host publicationMultimedia Technology IV - Proceedings of the 4th International Conference on Multimedia Technology
EditorsAly A. Farag, Jian Yang, Feng Jiao
PublisherCRC Press/Balkema
Pages53-57
Number of pages5
ISBN (Electronic)9781138027947
DOIs
StatePublished - 2015
Event4th International Conference on Multimedia Technology, ICMT 2015 - Sydney, Australia
Duration: 28 Mar 201529 Mar 2015

Publication series

NameMultimedia Technology IV - Proceedings of the 4th International Conference on Multimedia Technology

Conference

Conference4th International Conference on Multimedia Technology, ICMT 2015
Country/TerritoryAustralia
CitySydney
Period28/03/1529/03/15

Keywords

  • Image stitching algorithms
  • Smoothly varying affine stitching (SVAS)
  • Stability

Fingerprint

Dive into the research topics of 'Fast and robust varying affine stitching by proper feature point selection'. Together they form a unique fingerprint.

Cite this