BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A Two-stage Image Segmentation Method using a Convex Variant of th
 e Mumford-Shah Model and Thresholding - Raymond Chan\, The Chinese Univers
 ity of Hong Kong
DTSTART:20150703T140000Z
DTEND:20150703T150000Z
UID:TALK59464@talks.cam.ac.uk
CONTACT:Dr Jan Lellmann
DESCRIPTION:The Mumford-Shah model is one of the most important image segm
 entation models\, and\nhas been studied extensively in the last twenty yea
 rs. In this\ntalk\, we propose a two-stage segmentation method based on th
 e Mumford-Shah model.\nThe first stage of our method is to find a smooth s
 olution\ng to a convex variant of the Mumford-Shah model. Once g is obtain
 ed\,\nthen in the second stage\, the segmentation is done by thresholding 
 g\ninto different phases. The thresholds can be given by the users\nor can
  be obtained automatically using any clustering methods.\nBecause of the c
 onvexity of the model\,\ng can be solved efficiently by techniques like th
 e split-Bregman algorithm\nor the Chambolle-Pock method. We prove that our
  method is convergent\nand the solution g is always unique. In our method\
 , there is no need\nto specify the number of segments K before finding g.\
 nWe can obtain any K-phase segmentations by choosing K-1 thresholds\nafter
  g is found in the first stage\; and in the second stage\nthere is no need
  to recomputeg  if the thresholds are changed\nto reveal different segment
 ation features in the image.\nExperimental results show that our two-stage
  method performs better\nthan many standard two-phase or multi-phase segme
 ntation methods\nfor very general images\, including anti-mass\, tubular\,
  MRI\, noisy\, and blurry images\;\nand for very general noise models such
  as Gaussian\, Poisson and multiplicative\nGamma noise. We will also menti
 on the generalization to color images.
LOCATION:MR 14\, CMS
END:VEVENT
END:VCALENDAR