September 21, 2021
12:00 pm / 1:00 pm
Venue
https://wse.zoom.us/j/99567504456?pwd=WkI2UlpGT3p6MldLS05VNkdmcGxiZz09
?Three Recent Short Stories About ImageDenoising?
Mário Figueiredo, PhD
Distinguished Professor
University of Lisbon
Portugal
Abstract: In this talk, I will review threerecent proposals in thearea of patch-based image denoising. The first one is a simple post-processingtechnique for Poisson denoisers, based on an elementary application ofclassical linear minimum mean squared error (LMMSE) estimation, and which isable to squeeze a few extra tenths of dB of ISNR from several state-of-the-artPoisson denoisers and to produce better-lookingimages. The second story isabout how external non-local means (NLM) denoising can be seen as an importancesampling approach to computing MMSE patch estimates, which opens the door tousing NLM with arbitrary noise models. The third and final part of the talk isabout denoising interferometric (phase) images using multi-resolution windowed Fourierfiltering, guided by Stein’s unbiased risk estimate (SURE), outperformingprevious state-of-the-art methods for this problem.
Biography: MárioFigueiredo is a DistinguishedProfessor at Instituto Superior Técnico, University of Lisbon, with theDepartment of Electrical and Computer Engineering. He is also a seniorresearcher and group leader at Instituto de Telecomunicações. His research areasinclude machine learning, signal processing, and optimization. He has receivedseveral honors and awards, namely: Fellowof the Institute of Electricaland Electronics Engineers (IEEE), Fellow of the International Association forPattern Recognition (IAPR), Fellow of the European Association for SignalProcessing (EURASIP), IEEE Signal Processing Society Best Paper Award, W. R. G.Baker Award (IEEE), Pierre Devijver Award (IAPR), Technical Achievement Award(EURASIP), member of thePortuguese Academy of Engineering, member of theLisbon Academy of Science. From 2014 to 2018 he was included in the annual list?Highly Cited Researchers.?