@book{Magnor:2001:GAM,
                  author={M.~Magnor},
                  title={Geometry-Adaptive Multi-View Coding
                  Techniques for Image-based Rendering},
                  year=2001,
                  publisher={Shaker Verlag, Aachen, Germany},
                  address={ISBN 3-8265-8315-9},
		  note={Ph.D. Thesis, University Erlangen-Nuremberg,
		  Germany, Nov. 2000},
		  url="http://graphics.stanford.edu/~magnor/Publications.html",
		  annote="Photo-realistic rendering of real-world
		  scenes is an important research objective in
		  computer graphics.  Current geometry-based rendering
		  approaches often yield artificial-looking rendered
		  views because global illumination effects are very
		  time-consuming to simulate.  Image-based Rendering
		  (IBR) techniques, on the other hand, attain
		  natural-appearing rendering results of arbitrary
		  scenes.  In IBR, rendering frame-rate is independent
		  of scene complexity.  Instead, computational load is
		  directly proportional to the number of rendered
		  pixels.  IBR methods require large amounts of image
		  data as input.  Up to gigabytes of data need to be
		  acquired, stored, transmitted and processed for a
		  single scene.  Conventional image coding techniques
		  offer compression factors 1 to 2 orders of magnitude
		  too low to be applicable to IBR.  Goal of this work
		  is the development of suitable coding schemes for
		  IBR.  As the calibrated image data used for IBR
		  depict static three-dimensional scenes from multiple
		  viewpoints, images differ due to disparity,
		  occlusions and non-Lambertian object surface
		  reflection characteristics. Four different
		  compression methods are presented for various image
		  data characteristics.  Coding bit-rate and desired
		  image reconstruction quality can be varied over a
		  wide range in all coding schemes.  The coders'
		  rate-distortion performance is evaluated using a
		  number of different image test data sets.  The
		  experimental results verify the coding schemes'
		  applicability to IBR.  At moderate image
		  reconstruction quality, all codecs attain
		  compression factors exceeding \mbox{1000:1}."
}