BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:PaLEnTIR: a Parametric Level Set-based Approach to Image Reconstru
 ction and Restoration - Misha Kilmer (Tufts University)
DTSTART:20230111T150000Z
DTEND:20230111T160000Z
UID:TALK194488@talks.cam.ac.uk
DESCRIPTION:Inverse problems are of significant interest across a broad ra
 nge of science and engineering applications. The primary objective in an i
 nverse problem is to extract the unknown composition and structure of a me
 dium based on a set of indirect observations which are related to the unkn
 own via a physical model. In many cases\, one seeks only the identificatio
 n and characterization of ``regions of interest'' (ROIs)\, such as cancero
 us tumors from Xray CT or diffuse optical data\, subsurface contaminants f
 rom hydrological data or buried objects from electromagnetic data. These p
 roblems are often solved by first forming an image and then post processin
 g to identify the ROIs. &nbsp\;Although this can be effective\, it is comp
 utationally expensive. Moreover\, for data limited problems\, the initial 
 image formation stage will require potentially complex regularization meth
 ods and user tuning to overcome the ill-posed nature of these problems.\nW
 e propose an alternative approach in the context of inverse image reconstr
 uction/restoration of piecewise constant objects whereby we parameterize t
 he image model itself. Our PaLEnTIR model is a significantly enhanced para
 metric level set image model.&nbsp\; Instead of solving for pixel/voxel va
 lues\, we optimize for a very small set of parameters and the regularizati
 on is built into the model itself.&nbsp\; &nbsp\;Given upper and lower bou
 nds on the contrast\, our approach can recover objects with any distributi
 on of contrasts and eliminates the need to know either the number of contr
 asts in a given scene or their values. &nbsp\;Our inversion algorithm ther
 efore optimizes for the model parameters while iteratively estimating the 
 necessary space-varying contrast limits.&nbsp\; We discuss the properties 
 of our PaLEnTIR model and the inversion algorithm and demonstrate the perf
 ormance on several 2D and 3D linear and non-linear inverse problems.\n&nbs
 p\;
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR