Underwater Image Stitching

  • Author / Creator
    Hojjati, Saeed
  • For years, panoramic image stitching has been an interesting problem for re- searchers. Several advances have been made in stitching images that are ac- quired outside of water, but the problem has been poorly explored for under- water images. Image stitching for underwater images can be used in numerous scientific applications in the fields of marine geology, archaeology, and biology that involve tasks such as the prospection of ancient shipwrecks, the detection of temporal changes, and environmental damage assessment. It can also be used to create virtual reality tours of zones of special interest such as under- water nature reserves. Underwater images suffer from poor visibility conditions because of medium scattering, light distortion, and inhomogeneous illumination. This causes many image stitching techniques to fail when they are applied to underwa- ter images. In this work, we develop a novel method for stitching underwater images. We adopt dehazing to not only improve the aesthetic quality of the images but also to enable feature detectors to accurately detect and match feature points. We also adopt guided image filtering to improve the speed of the dehazing algorithm. A novel idea proposed in this method is to use colour transfer to transform images into the same colour space in order to re- duce lighting inhomogeneities and exposure artifacts. We further process our stitched results by applying a graph-cut strategy that operates in the image gradient domain over the overlapping regions to improve blurring and ghosting effects caused by local misalignments. Finally, we apply a transition smooth- ing method to produce more plausible results and to reduce the noticeability of the joining regions to an even higher degree.

  • Subjects / Keywords
  • Graduation date
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
  • Language
  • Institution
    University of Alberta
  • Degree level
  • Department
    • Department of Computing Science
  • Supervisor / co-supervisor and their department(s)
    • Yang, Herb (Computing Science)
  • Examining committee members and their departments
    • Sander, Joerg (Computing Science)
    • Boulanger, Pierre (Computing Science)