Jonas August, PhD

Project Scientist

Healthcare Robotics Center

Robotics Institute

Carnegie Mellon University

 
 


Research

Geometry and uncertainty are coupled aspects of my interests in computer vision and medical imaging. On the one hand, we seek descriptions of the structures in the physical world in the language of differential geometry. On the other, our observations of this world come in the form of amorphous, noisy images, without the discrete structures we effortlessly call "objects".  I am interested in "organizing" images into such objects, and I use the tools of Markov processes and random fields to derive algorithms from first principles.  Not only are such ideas relevant for helping a roving robot deal with its environment or for understanding the neural architecture of the primary visual cortex in primates, but they can be exploited to improve medical image quality.  For example, I am exploring the use of probabilistic techniques for removing the often catastrophic "streaking artifacts" in X-ray CT scans.  In addition, by building in a model of the local geometry of curves, one can obtain vascular structures from low quality medical images.

Projects

Parallel methods in regularized tomography
"doc": a GNU/Linux cluster for medical imaging applications

Interests

  • Medical imaging, regularization, image artifact removal, mathematical modeling
  • Image statistics: What structure can be observed? Can it be modeled tractably?
  • Finding curves in images: inference of 1-d structures in 2-d images
  • Putting together curves to form objects: contour fragment grouping
  • Stably representing objects: Hierarchies, axial descriptions, "shape"
  • Doctoral Dissertation

    The Curve Indicator Random Field

    Selected Publications

  • ``The Role of Non-Overlap in Image Registration,'' in Proceedings, Information Processing in Medical Imaging (IPMI), 2005. Download paper in PDF.
  • ``Weakly-Supervised Segmentation of Non-Gaussian Images via Histogram Adaptation,'' in Proceedings, MICCAI 2003. Download paper in PDF.
  • ``Volterra Filtering of Noisy Images of Curves,''in Proceedings,  European Conference on Computer Vision, Copenhagen, May 2002. Download paper in PDF.
  • ``Decoupling the Equations of Regularized Tomography,''in Proceedings, 2002 IEEE International Symposium on Biomedical Imaging, Washington, D.C., July 7--10, 2002.  Download paper in PDF.
  • ``A Markov Process Using Curvature for Filtering Curve Images,'' with Steven Zucker, in Energy Minimization Methods in Computer Vision and Pattern Recognition, Nice, 2001. Download paper in gzipped PostScript or PDF.
  • ``A Generative Model of Curve Images with a Completely-Characterized Non-Gaussian Joint Distribution,'' with Steven Zucker, in Workshop on Statistical and Computational Theories of Vision at International Conference on Computer Vision, July 2001. Download paper in gzipped PostScript or PDF.
  • ``A Field Model for Contour Organization and Partial Differential Equations,'' with Steven Zucker, in Workshop on Perceptual Organization in Computer Vision at International Conference on Computer Vision, July 2001. Download paper in gzipped PostScript or PDF.
  • ``The Curve Indicator Random Field: Curve Organization via Edge Correlation,'' with Steven Zucker, Chapter. 15, pp. 265-287, in Perceptual Organization in Artificial Vision Systems, Boyer and Sarkar, Eds., Kluwer Academic, Boston, 2000. Download paper in gzipped PostScript.
  • ``On the Evolution of the Skeleton,'' in IEEE International Conference on Computer Vision, September 1999. Download paper in gzipped PostScript. Movies (animated GIFs) of evolving objects and their skeletons: handhand skeleton, pimple, semiligature, branch lengthening despite boundary shortening, junction switch (an "internal" skeleton instability is flagged), birth
  • ``Ligature Instabilities and the Perceptual Organization of Shape,'' with Kaleem Siddiqi and Steven Zucker,Computer Vision and Image Understanding 76(3): 231-243, 1999. Download in gzipped PostScript.
  • ``Contour Fragment Grouping and Shared, Simple Occluders,'' with Kaleem Siddiqi and Steven Zucker, Computer Vision and Image Understanding, 76(2): 146-162, 1999. Download in gzipped PostScript.
  • ``Duality in Perceptual Organization: Grouping vs. Shape Decomposition,'' in IEEE Workshop in Perceptual Organization in Computer Vision, June 1998.
  • ``Fragment grouping via the principle of perceptual occlusion,'' inInternational Conference on Pattern Recognition, August 1996.
  • Education

    Resume / Curriculum Vitae

    Download in PDF.      

    Friends and Colleagues

    Ohad Ben-Shahar
    Athos Georghiades
    Patrick Huggins
    Takeo Kanade
    Ben Kimia
    David Mumford
    Kaleem Siddiqi
    Lance Williams
    Song-Chun Zhu
    Steve Zucker
     
     
     

     /etc
     
     

    Contact:

    Jonas August
    The Robotics Institute
    Carnegie Mellon University
    5000 Forbes Avenue
    Pittsburgh, PA 15213-3890

    Tel: (412) 268-1314
    FAX: (412) 268-6436
    Email: jonas@cs.cmu.edu