BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Cambridge Image Analysis Seminars
SUMMARY:Modular Optimization Strategies for Compression an
 d Restoration of Visual Data - Yehuda Dar (Technio
 n\, Haifa)
DTSTART;TZID=Europe/London:20190520T150000
DTEND;TZID=Europe/London:20190520T160000
UID:TALK125197AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/125197
DESCRIPTION:Many important imaging problems have complex struc
 tures that challenge the design of their algorithm
 ic solutions. Contemporary methods for addressing 
 such intricate problems are often formed as concep
 tually-simple iterative procedures that rely on we
 ll-engineered black-box modules\, usually in the f
 orm of existing denoising techniques. \nThis talk 
 overviews an alternative optimization framework to
  address intricate imaging problems based on promi
 nent compression standards as the fundamental modu
 les iteratively applied. The proposed algorithms u
 tilize the alternating direction method of multipl
 iers (ADMM) optimization technique to address seem
 ingly impractical tasks\, involving aspects of acq
 uisition\, compression and rendering of images. Sp
 ecifically\, we demonstrate how the state-of-the-a
 rt HEVC compression standard can be adjusted to se
 veral layouts of acquisition and rendering systems
 \, such that the end-to-end reconstruction errors 
 of the systems are optimized with respect to the c
 ompression bit-costs. The experiments established 
 our method as the best approach for optimizing the
  system performance at high bit-rates from the com
 pression standpoint.\nMoreover\, we relate the pro
 posed approach also to image restoration using com
 plexity regularization\, where the likelihood of c
 andidate solutions is evaluated based on their com
 pression bit-costs. Our proposed methods restore s
 ignals by leveraging state-of-the-art models desig
 ned for compression. The presented experiments sho
 w good results for image deblurring and inpainting
  using the JPEG2000 and HEVC compression standards
 .\n\nThis is joint work with Prof. Alfred Bruckste
 in and Prof. Michael Elad.
LOCATION:MR 4\, Centre for Mathematical Sciences
CONTACT:Carola-Bibiane Schoenlieb
END:VEVENT
END:VCALENDAR