BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:3D Reconstruction meets GPGPU meets Image Analysis - Christopher Z
 ach
DTSTART:20110408T090000Z
DTEND:20110408T100000Z
UID:TALK30577@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:In this talk I give an overview of my recent research activiti
 es in computer\nvision during my post-doctoral stays at UNC-Chapel Hill an
 d ETH Zurich. The\nprimary goal of generating faithful virtual representat
 ions of real\nenvironments from image data led to research in two methodol
 ogically very\ndifferent fields: first\, I investigated in the structure-a
 nd-motion problem\nfor large image collections.  A lot of my research on t
 his topic addresses the\nquestion of how difficult and ambiguous visual st
 ructures--potentially leading\nto non-recoverable incorrect 3D models--can
  be detected. The proposed\nprobabilistic models go beyond the standard re
 projection error to assess the\nquality of the structure-and-motion soluti
 on. The demand for 3D mesh models\nfrom images with optimality guarantees 
 initiated a second line of research\nrooted in continuous optimization. Th
 e proposed\, relatively simple energy\nformulation for surface reconstruct
 ion yields high-quality results even for\ninputs strongly contaminated wit
 h noise and outliers. The utilized methodology\nalso extends to different 
 application areas such as medical imaging and fast\ninference in graphical
  models. Efficiency of these methods is assured by\nutilizing graphics pro
 cessing units.\n
LOCATION:Small lecture theatre\, Microsoft Research Ltd\, 7 J J Thomson Av
 enue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR