BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Monday Seminar : Image Compression with Differential Equations
DTSTART:20111003T161500
DTSTAMP:20260501T144225Z
UID:412d078ef5e7ca8a9de5a596e7c4598efb54b4ce658474b31c1cc216
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Joachim Weickert - Saarland University\, Germany - Invit
 ed by Prof. Pascal Fua\nAbstract One of the most fascinating aspects of pa
 rtial differential equations (PDEs) in image processing is their filling-i
 n effect. It allows to recover missing image structures by interpolating d
 ata from the neighbourhood. In this talk we exploit this property for loss
 y compression of digital images. The idea is to keep only a small fraction
  of the pixels and to reconstruct the remaining data with PDE-based interp
 olation. This gives rise to three interdependent questions: 1. Which are t
 he best pixels for being kept? 2. How can the selected pixels be encoded i
 n an efficient way? 3. What are the most useful PDEs for data interpolatio
 n? In order to gain insights into the pixel selection problem we first foc
 us on interpolation by homogeneous diffusion\, where we present theoretica
 l results\, optimization strategies and fast algorithms. To encode the sel
 ected data efficiently\, we use contour data or tree-based representations
 . Afterwards we evaluate various PDE-based interpolation operators and sho
 w why anisotropic nonlinear diffusion has very favourable properties. It i
 s used to design a codec that can outperform the JPEG standard and even it
 s more advanced wavelet-based successor JPEG 2000.
LOCATION:INM 202
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR