Applications of Hardware-Accelerated Filtering in Computer Graphics

Information

Abstract

Two of the most important issues of computer graphics – especially in raster graphics and volume visualisation – are sampling and reconstruction. These operations must fulfill particular conditions of sampling theory in order to be able to represent arbitrary continuous functions by discrete samples and reconstruct them from these samples without significant information loss. Several approaches for high-quality reconstruction have been introduced to the computer graphics community. These approaches are mostly implemented in software, possible only as pre-process step. The only two ways of reconstruction that are usually fast enough for real-time rendering are nearest neighbor and linear interpolation filtering, but the quality of these filtering processes is often not sufficient. This work summarizes the hardware-based methods that exploit the features of today’s graphics chips for filtering tasks. These methods are divided in two parts, i.e., high-resolution filtering and image processing. Both methods are based on the distribution principle of convolution known from splatting based volume rendering algorithms and they share the same general principle. The difference is in the implementation of the algorithms themselves. High-resolution filtering employs high-order filters in order to improve the quality of resampling tremendously. The implemented algorithms exploit symmetry or separability properties to make the filtering more efficient. We compare it to the existing, natively supported solution, i.e., linear interpolation to our filtering implementation using higher order filters. This is shown in various application areas like surface-texturing, and solid-texturing, animated textures, or derivative filtering; at interactive framerates. The image processing algorithms are simplified general filtering algorithms to increase effciency and performance. We show the usability on smoothing and edge detection, the important operations of image processing and pattern recognition. We combine this techniques together with other hardware features to provide hardware-accelerated artistic rendering techniques. These are also presented in a rendering system that provides non-photorealistic rendering effects.

Additional Files and Images

Additional images and videos

Additional files

Weblinks

No further information available.

BibTeX

@mastersthesis{Masterthesis-Viola,
  title =      "Applications of Hardware-Accelerated Filtering in Computer
               Graphics",
  author =     "Ivan Viola",
  year =       "2002",
  abstract =   "Two of the most important issues of computer graphics –
               especially in raster graphics and volume visualisation –
               are sampling and reconstruction. These operations must
               fulfill particular conditions of sampling theory in order to
               be able to represent arbitrary continuous functions by
               discrete samples and reconstruct them from these samples
               without significant information loss. Several approaches for
               high-quality reconstruction have been introduced to the
               computer graphics community. These approaches are mostly
               implemented in software, possible only as pre-process step.
               The only two ways of reconstruction that are usually fast
               enough for real-time rendering are nearest neighbor and
               linear interpolation filtering, but the quality of these
               filtering processes is often not sufficient. This work
               summarizes the hardware-based methods that exploit the
               features of today’s graphics chips for filtering tasks.
               These methods are divided in two parts, i.e.,
               high-resolution filtering and image processing. Both methods
               are based on the distribution principle of convolution known
               from splatting based volume rendering algorithms and they
               share the same general principle. The difference is in the
               implementation of the algorithms themselves. High-resolution
               filtering employs high-order filters in order to improve the
               quality of resampling tremendously. The implemented
               algorithms exploit symmetry or separability properties to
               make the filtering more efficient. We compare it to the
               existing, natively supported solution, i.e., linear
               interpolation to our filtering implementation using higher
               order filters. This is shown in various application areas
               like surface-texturing, and solid-texturing, animated
               textures, or derivative filtering; at interactive
               framerates. The image processing algorithms are simplified
               general filtering algorithms to increase effciency and
               performance. We show the usability on smoothing and edge
               detection, the important operations of image processing and
               pattern recognition. We combine this techniques together
               with other hardware features to provide hardware-accelerated
               artistic rendering techniques. These are also presented in a
               rendering system that provides non-photorealistic rendering
               effects.",
  month =      apr,
  note =       "1",
  address =    "Favoritenstrasse 9-11/E193-02, A-1040 Vienna, Austria",
  school =     "Institute of Computer Graphics and Algorithms, Vienna
               University of Technology ",
  URL =        "/research/publications/2002/Masterthesis-Viola/",
}