image processing: adapting keypoint matching across domains

Share

   Image processing is fundamental to research in areas such as neutron imaging, computer vision, medical imaging, and remote sensing. However, while many core image processing concepts are transferable across  domains, appropriate implementation may require unique modifications to adapt the algorithms to the domain phenomenology.

    In this talk I will focus on how feature-based image registration methods differ for various applications, and how they can be adapted for optimal performance. I will describe the concept of control point or “keypoint” matching from a general perspective. I will first discuss its implementation in the computer vision domain. I will then show the limitations of applying this implementation directly to applications in remote sensing. Based on these limitations, a second implementation will then be proposed from the remote sensing point of view, and conclusions will be drawn related to algorithm transferability. This talk will build an intuition for adapting generic image processing techniques to specific domains of interest.



  Date and Time

  Location

  Contact

  Registration


  • Rochester Institute of Technology
  • 54 Lomb Memorial Drive
  • Rochester, New York
  • United States 14623
  • Building: Center for Imaging Science Bld 76
  • Room Number: 1275
  • Click here for Map


  Speakers

Sophie Voisin
Sophie Voisin of Oak Ridge National Lab

Topic:

image processing: adapting keypoint matching across domains

Image processing is fundamental to research in areas such as neutron imaging, computer vision, medical imaging, and remote sensing. However, while many core image processing concepts are transferable across  domains, appropriate implementation may require unique modifications to adapt the algorithms to the domain phenomenology.

    In this talk I will focus on how feature-based image registration methods differ for various applications, and how they can be adapted for optimal performance. I will describe the concept of control point or “keypoint” matching from a general perspective. I will first discuss its implementation in the computer vision domain. I will then show the limitations of applying this implementation directly to applications in remote sensing. Based on these limitations, a second implementation will then be proposed from the remote sensing point of view, and conclusions will be drawn related to algorithm transferability. This talk will build an intuition for adapting generic image processing techniques to specific domains of interest.

Biography:

Dr. Sophie Voisin is a R&D associate at ORNL developing high performance computing methods for geospatial data analysis for the GIST group.  She received her PhD in Computer Science and Image Processing from the Université de Bourgogne (France) in 2008 and joined ORNL in 2010 to work on numerous image processing related projects, successively performing quantitative analysis of neutron 2D and 3D image data; developing new techniques for eye-gaze data analysis, for which she is a co-recipient of an R&D 100 award (2014); and now implementing multidimensional image processing algorithms on GPU platforms for high performance computing of satellite imagery.

Email:

Address:Oak Ridge National Lab, , Oak Ridge, New York, United States, 37830