BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Dealing with damage: Region-based image segmentation despite artif
 acts and bias 
DTSTART:20150618T103000
DTSTAMP:20260407T024532Z
UID:854504c93e5cb1244fbc173ee020259e7f40fa5997b81ffb42ff127c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr Dominique Zosso\, University of California\, Los Angeles\nB
 io: Dominique Zosso received the M.Sc. degree in electrical and electronic
 s engineering and the Ph.D. degree from EPFL in 2006 and 2011\, respective
 ly.\nHe was Research and Teaching Assistant at the Signal Processing Labor
 atory LTS5\, EPFL\, from 2007 to 2012. He then became SNSF postdoctoral fe
 llow\, and is currently a CAM Assistant Adjunct Professor with the Departm
 ent of Mathematics\, University of California\, Los Angeles\, CA\, USA\, w
 orking with Luminita A. Vese\, Andrea L. Bertozzi and Stanley J. Osher. Hi
 s research interests include variational and PDE methods\, and efficient a
 lgorithms to solve inverse problems in imaging and computer vision\, and r
 elated problems in machine learning and data science. He develops mathemat
 ical models and cutting edge algorithms\, to help understand images and so
 lve problems in medicine\, biology\, robotics and consumer electronics.\nR
 egion-based image segmentation has essentially been solved by the Chan-Ves
 e (CV) model. However\, this model fails when images are affected by artif
 acts (outliers) and illumination bias that outweigh the actual image contr
 ast. Here\, we introduce a model for segmenting such images. In a single e
 nergy functional\, we introduce 1) a dynamic artifact class preventing int
 ensity outliers from skewing the segmentation\, and 2)\, in Retinex-fashio
 n\, we decompose the image into a piecewise-constant structural part and a
  smooth bias part. The CV-segmentation terms then only act on the structur
 e\, and only in regions not identified as artifacts. The segmentation is p
 arameterized using a phase-field\, and efficiently minimized using thresho
 ld dynamics.\nWe demonstrate the proposed model on a series of sample imag
 es from diverse modalities exhibiting artifacts and/or bias. Our algorithm
  typically converges within 10-50 iterations and takes fractions of a seco
 nd on standard equipment to produce meaningful results. We expect our meth
 od to be useful where image damage prevents classical CV-segmentation from
  working\, and anticipate use in applications where artifacts and bias are
  actual features of interest\, such as lesion detection and bias field cor
 rection in medical imaging\, e.g. in magnetic resonance imaging (MRI).
LOCATION:DIA003
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR