BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stylized Rendering for Multi-resolution Image Representation: Reco
 nciling Efficiency with Aesthetics in Image Compression - Neil Dodgson\, C
 omputer Laboratory
DTSTART:20080507T131500Z
DTEND:20080507T141500Z
UID:TALK11992@talks.cam.ac.uk
CONTACT:Timothy G. Griffin
DESCRIPTION:\n\nWe integrate stylized rendering with an efficient multi-re
 solution image\nrepresentation. This enables the user to control how compr
 ession affects\nthe aesthetic appearance of an image. We use a point-based
  rendering\napproach for progressive image transmission and compression. T
 his\nrepresents an image by a sequence of colour values. To best approxima
 te\nthe image at progressive levels of detail\, we developed a novel\,\nad
 aptive farthest point sampling algorithm that balances global coverage\nwi
 th local precision. We store no spatial information other than the\naspect
  ratio: the spatial position of each colour value is inferred from\nthe pr
 eceding members of the sampling sequence. We keep track of the\nspatial in
 fluence of each sample on the rendition\, and use a\nprogressively-generat
 ed discrete Voronoi diagram as the common\nfoundation for our sampling and
  rendering framework. This framework\nallows us to extend traditional phot
 orealistic methods of image\nreconstruction by scattered data interpolatio
 n to encompass non-\nphotorealistic rendering. It supports a wide variety 
 of artistic\nrendering styles based on geometric subdivision or parametric
  procedural\ntextures. We compare our results with standard JPEG compressi
 on and\nconclude with a brief overview of the implications of using\nnon-p
 hotorealistic representations for highly compressed imagery. This\nwork wa
 s undertaken largely by Dr Mark Grundland (Dr Dodgson's PhD\nstudent). Par
 t of the work formed a Part II project undertaken by Mr\nChris Gibbs.
LOCATION:Lecture Theatre 1\, Computer Laboratory
END:VEVENT
END:VCALENDAR